Roma poverty from a human development perspective

http://europeandcis.undp.org/ourwork/roma/

Roma Inclusion Working Papers

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Roma Inclusion Working Papers UNDP Regional Support Centre for Europe and CIS, Istanbul

This publication was prepared with the support from the SDC, Swiss Development Cooperation. It is the latest in a series, the previous contributions of which were financed by the European Commission’s Directorate General for Regional and Urban Policy

UNDP, 2014

ISBN (printed version): 978-92-95092-89-1

To be cited as: Ivanov, A., Kagin, J. (2014). Roma poverty from a human development perspective. Roma Inclusion Working Papers. Istanbul: UNDP.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted, in any form or by any means, mechanical, photocopying, recording or otherwise without prior written permission of the publishers.

The views expressed in this publication should in no way be taken as reflecting the views of the UNDP, SDC or the European Union.

Cover design and layout: Yassen Panov Print: Print BG

Editors of the Roma Inclusion Working Papers: Ben Slay and Jasmina Papa.

About the authors Dr. Andrey Ivanov has a Masters Degree in Economics from the Central School of Commerce, Warsaw, and a Ph.D. in Modern History from the Institute of International Economics and Political Studies, Moscow. At the time of writing this paper he was a Senior Policy Advisor in the UNDP Regional Centre in Bratislava where he coordinated analytical work and quantitative data collection on human development, social exclusion and the Roma. Dr. Ivanov also served as the lead author of a number of regional reports published by UNDP on these topics. He is now heading the ‘Roma and Migrants Integration’ sector of the European Union Agency for Fundamental Rights; the analysis and views contained in this paper do not necessarily reflect the official position of the Agency.

Dr. Justin Kagin has a Ph.D. in agricultural and resource economics from the University of California, Davis. He has worked on various projects with UNDP for the past 6 years, specifically on issues concerning the Roma, social exclusion and human development. He is currently the owner and founder of Kagin’s Consulting, a poverty alleviation development company.

Acknowledgements This paper reflects the collective work and discussions of many people involved in Roma-related research done by UNDP Regional Centre in Bratislava in the period 20022013. So the list of acknowledgements would be inevitably incomplete. However it is fair to start it with Kalman Mizsei, former director of the Regional Bureau for Europe and CIS of UNDP, who actually launched UNDP’s involvement in Roma inclusion. Ben Slay and Jens Wandel, the two (also former) directors of the Regional Centre in Bratislava were instrumental in supporting this research, both managerially and conceptually. Special thanks are due to Balázs Horvath, the (also former) head of the Poverty Practice of BRC, where the work on sustainable human development, social inclusion and Roma vulnerability was conducted. Even while in South Sudan as a UNDP Country Director he continued to comment on drafts and shared ideas. Special thanks are due to the members of the “Roma inclusion team” at BRC, which throughout the years included Jaroslav Kling, Susanne Milcher and Daniel Škobla. The work on this paper was supported, morally and conceptually, by a number of colleagues and friends. Sabina Alkire was instrumental in fine tuning the multidimensional poverty index to the specific context of Roma poverty. Ilona Tomova provided valuable insights on the qualitative aspects of the quantitative data and on the diversity of survival strategies. Monica Rossi alerted us to the overlaps between the classical classbased tensions and the horizontal inequalities in Roma communities. Dorothy Rosenberg copy edited the text and streamlined its core messages. Adib Nehmeh helped address few remaining inconsistencies in the final version. Ioannis Dimitrakopoulos was particularly supportive in positioning the analysis in a broader policy context. The authors of the other publications in the “Roma Inclusion Working Papers” series were implicitly and explicitly part of the creative process. They were involved in a genuine analytical debate on the nature and determinants of poverty and exclusion of the Roma; some of them (namely Dotcho Mihailov and Tatjana Perić) also commented on different stages of drafting this paper. Jelena Tadzic and Jasmina Papa contributed to the finalization of the draft with last-minute comments and ideas. Finally, thank you to J. Edward Taylor, Travis Lybbert and Richard Green from the University of California, Davis who provided valuable comments to the econometrics models used and the basis with which to construct an appropriate index. If any flaws have remained after such massive support was received, they should be attributed entirely to the authors.

4

Executive summary Reducing Roma poverty was placed on the national political agenda and formulated as an explicit commitment of the countries participating in the Decade of Roma Inclusion (20052015). The European Commission later joined in with its EU Framework of Roma Inclusion, putting additional political weight behind the effort to lift Roma out of poverty. But what poverty are we talking about? What is “poverty” is apparently an easy question but its simplicity is misleading. This apparent simplicity often drives researchers, activists, and policy makers towards equally simplistic schemes for addressing the issue. This paper goes beyond the standard (still prevailing) approaches to poverty in terms of poverty rates. It presents the basic concepts and approaches to the definition of poverty; applies them to the Roma using the data on the status of Roma households from the surveys conducted by UNDP, the World Bank and the European Commission; and proposes a multidimensional measure that better reflects the multidimensional challenges Roma are facing. The data is drawn from the 2011 survey, in conjunction with the FRA pilot Roma survey conducted during the same period, and represents the largest integrated household survey of the Roma to date. The combined UNDP/WB/EC regional Roma survey and the FRA Roma pilot survey (both from 2011) has data produced by 20,018 Roma households (with 87,717 household members living in these households) and 9,782 non-Roma households living nearby (28,214 household members), covering 18 European countries. The 2011 survey followed the conceptual approach tested in 2004 by UNDP in its first comprehensive regional survey of Roma at risk of marginalization and non-Roma living in close proximity (UNDP 2006).1 Using this data, the paper analyses standard monetary poverty measures for the Roma but then quickly proceeds to describe more non-pecuniary measures. Education, employment, health and housing are analysed in depth showing that the nature of Roma poverty and status vis-à-vis the majority is more than just a matter of income. A full analysis must combine these various dimensions and thus, a multi-dimensional index is constructed from these critical areas plus indicators from a rights-based approach. This index is compared to monetary measures and applied to other indicators of exclusion. The key results from the application of the different poverty measures are given below in this summary.

Uni-dimensional poverty measures applied to the Roma z The two approaches to calculating monetary poverty (by income and by expenditures) yield a generally consistent picture. Income estimates usually produce higher poverty rates (people tend to underreport income)—and this is also the case of both datasets, of the Roma and non-Roma. In all countries with the exception of Bosnia and Herzegovina both approaches show similar trends of declining poverty. z Access to early childhood education and care appears to be a major challenge for Roma—particularly as non-attendance of pre-school clearly limits professional 1/ The 2004 survey was prepared and implemented by UNDP. The 2011 survey was implemented jointly by UNDP, the World Bank and the European Commission, Directorate General “Regional and Urban Policy”.

5

and educational opportunities later in life. Roma were much less likely to attend pre-school than non-Roma living in close proximity. Differences in pre-school attendance rates between Roma and non-Roma were statistically significant in all countries surveyed, except for Bosnia and Herzegovina, FYR Macedonia, and Montenegro (countries with the lowest overall pre-school attendance rates). z Unsurprisingly, throughout Central and Southeast Europe, the Roma face higher unemployment rates than non-Roma populations living in their close proximity. Roma are more likely to be unemployed than their non-Roma counterparts in all countries and for both men and women. Also, the relation to the national averages of both Roma and non-Roma populations living in close proximity varies across countries. This in part reflects the geographical distribution of vulnerable Roma communities which, for example, in Slovakia and Hungary, are concentrated in more impoverished parts of the country. Thus, both Roma and nonRoma unemployment rates based on the UNDP/WB/EC regional Roma survey are significantly higher than the national averages. z Financial access to (affordability of ) health services is also significantly lower among the Roma. At the regional level (unweighted averages of pooled Roma and non-Roma samples) 55% of the Roma report instances in the past 12 months when household members could not afford purchasing prescribed medicines compared to 25% of the non-Roma. On the other hand, reported access to primary health services (insurance and access to general practitioners) has significantly improved among the non-Roma during 2004-2011; it declined for Roma in Albania, Romania, and Bulgaria. The loss of access to primary health care is particularly striking in Romania: whereas 90% of Roma survey respondents reported having a family doctor in 2004, only 49%, reported being insured in 2011. z The differences in regards to access to basic infrastructures within the individual countries are also revealing. The biggest problem in all countries is access to sanitation. In some cases (Hungary, or Slovakia) limited access to improved water sources is matched by equal limitations in access to sanitation. In most countries, however, this is not the case. In Bulgaria and FYR Macedonia, for example, the share of those without indoor toilets or bathrooms is respectively 12 and 11 times higher than the share of those without piped water. This could reflect the general underdevelopment of communal infrastructures in rural areas (where significant numbers of Roma in those countries live).

Multi-dimensional poverty analysis The multidimensional poverty approach is particularly appropriate for addressing the issue of Roma poverty, precisely because of the multidimensional nature of the deprivation and marginalization that Roma are facing. Roma poverty is not just a lack of financial resources, unemployment, sub-standard housing, or poor access to social services. It is about a combination of all these factors, which are both outcomes of past spells of exclusion and determinants of future deprivations—reinforcing the vicious circle of poverty. In order to capture these aspects of Roma deprivation, a multidimensional poverty index has been developed and tested following the standard Alkire and Foster (2007) methodol-

6

ogy. This index integrates important aspects of human poverty reflecting appropriately the specifics of Roma exclusion. This index reflects the status of the individuals (with their characteristics) living in households (with their characteristics) and facing certain numbers of deprivations. It integrates 12 equally weighted indicators reflecting status in six critical dimensions from a human development perspective (basic rights, health, education, housing, standard of living, and employment). The results show that the multidimensional poverty rate of Roma has decreased substantially between 2004 and 2011 in most countries (it increased in Albania and Serbia). This is clearly an achievement. In most of those countries witnessing improvements, however, they resulted primarily from declines in the number of Roma in the “poor” category (5-7 deprivations). “Severe poverty” (more than 7 deprivations) decreased significantly only in Bosnia and Herzegovina and Romania; in Albania, it even increased. The choice of the metrics indeed matters. The different poverty measures yields different results. The different outcomes resulting from the application of multidimensional versus monetary metrics are analysed for the period between 2004 and 2011. The most drastic is the case of Albania where the monetary poverty rate of Roma was more than cut in half (from 78% to 36%) while multidimensional poverty rate increased from 49% to 66%. Likewise, in Serbia monetary poverty declined from 57% to 26%, while the multidimensional poverty rate increased from 51% to 55%. In other countries the direction of the change was the same (towards a decline in poverty measured by both metrics) but the magnitude of change differs in some countries. Multidimensional poverty rates among Roma declined more sharply than monetary poverty in Bosnia and Herzegovina, Montenegro, and FYR Macedonia; the decline was similar in both metrics in Croatia and Bulgaria; while the decline in monetary poverty in Romania was stronger than in multidimensional poverty. Probits and a poverty policy analysis are explored to reveal what might decrease the MPI. And finally, other correlations with the MPI are evaluated including measures of vulnerability across dimensions and behaviours of the different groups.

Conclusions Based on these findings, the paper concludes that making the right choice among the wide range of methodologies for measuring poverty is not easy. Actually, there is no “right” or “wrong” choice. The choice of the particular approach should be guided by policy priorities, not by pure research curiosity. The ultimate purpose is to better understand the underlying mechanisms and the factors determining poverty—and to help overcome them. The specifics of Roma poverty require multidimensional approaches. The proposed multidimensional poverty index is one example of such an approach. It is better suited for tracking poverty of marginalized and excluded groups like Roma and can be applied with negligible additional cost. However it is also not perfect because a major dimension of human development—that of agency—is missing, both due to conceptual reasons and data deficits. Addressing these gaps would make possible developing a comprehensive truly human development centred poverty measure. The analyses presented highlight the differences between money- and people-centred measures of poverty. They are important to bear in mind not just (and not primarily) for proper poverty monitoring, but also for policy formulation. The conceptual framework behind the poverty analysis determines the framework of responses.

7

Contents About the authors ............................................................................................................................. 3 Acknowledgements ........................................................................................................................ 4 Executive summary........................................................................................................................... 5 Contents ...................................................................................................................... 8 List of tables ................................................................................................................ 9 List of figures ............................................................................................................ 10 Introduction .............................................................................................................. 13 Various approaches to poverty estimation ............................................................................14 Institutional clustering of poverty monitoring methods .................................................16 The conceptual framework behind this analysis .................................................................19 The sources and the constraints of this research .................................................................19 Looking at poverty from a monetary perspective ................................................. 22 Absolute poverty profiles .............................................................................................................23 Relative poverty profiles ...............................................................................................................24 Income and expenditures ............................................................................................................29 Poverty from the perspective of final outcomes.................................................... 34 Educationally poor ..........................................................................................................................34 Employment poverty .....................................................................................................................39 Health ..................................................................................................................................................46 Housing ...............................................................................................................................................49 Integrated proxy for outcome-level poverty.........................................................................52 Looking beyond material wellbeing ...................................................................... 54 From “basic needs” to “human development”.......................................................................54 From “basic needs” to social exclusion.....................................................................................55 From “Human Development” to “Multidimensional Poverty” .........................................57 Roma poverty from a multidimensional perspective............................................ 58 The construction of the index.....................................................................................................58 Drivers of multidimensional poverty .......................................................................................63

8

Correlates with other dimensions of poverty .......................................................................67 Education ..............................................................................................................................68 Employment ........................................................................................................................71 Behavioural patterns and aspirations ......................................................................................73 The missing dimension of agency.............................................................................................77 Comparing the different approaches ..................................................................... 81 Multidimensional versus money-metric measures .............................................................81 Other estimates of well-being ....................................................................................................84 Conclusions .............................................................................................................. 86 Annexes ..................................................................................................................... 88 Bibliography ......................................................................................................................................88 Contry profiles ..................................................................................................................................92 Other publications in the “Roma Inclusion Working Papers” series ..............................96

List of tables Table 1: Different approaches to poverty ......................................................................................17 Table 2: Dimensions and indicators of the “Roma multidimensional poverty index” ...59 Table 3: Average number of deprivations .....................................................................................62 Table 4: Individual characteristics’ contribution to the risk of poverty ...............................66 Table 5: Average length of unemployment for those that have worked before and are currently unemployed (as per the ILO definition) in the age group 15-64 .........................................................................................................................71

9

List of figures Figure 1: Absolute monetary poverty (expenditure based) of Roma and non-Roma, 2004 and 2011(%).................................................................................22 Figure 2: Absolute monetary poverty (income based) of Roma and non-Roma, 2004 and 2011 (%)................................................................................................................23 Figure 3: Monetary poverty and severe poverty rates (expenditure based) and poverty gap for Roma, 2004 and 2011 (%) .........................................................25 Figure 4: Monetary poverty and severe poverty rates (income based) and poverty gap for Roma, 2004 and 2011.................................................................25 Figure 5: Different dimensions of inequality: relative poverty rates for Roma, their non-Roma neighbours and the country averages (selected countries), 2011 (%) ..26 Figure 6: Magnitude of horizontal inequality: Gini coefficients for Roma, their non Roma neighbours and the national averages (selected countries), 2011 (%) .......27 Figure 7: Structure of Roma household income, 2011 (%) ......................................................28 Figure 8: Structure of non-Roma household income, 2011 ....................................................28 Figure 9: Structure of Roma household expenditures, 2011 ..................................................30 Figure 10: Structure of non-Roma household expenditures, 2011 ......................................30 Figure 11: Structure of Roma and Non-Roma household income by poverty status, 2011 ...................................................................................................31 Figure 12: Structure of Roma and non-Roma household expenditures by poverty status, 2011 ...................................................................................................32 Figure 13: Pre-school attendance (%) .............................................................................................35 Figure 14: Roma with at least lower secondary education (%) ..............................................36 Figure 15: Early leavers from education and training (%) ........................................................37 Figure 16: Ethnically segregated schools (%) ...............................................................................38 Figure 17: Unemployment rates for Roma and non-Roma in Central and Southeast Europe, 2011 (%) ..............................................................40 Figure 18: Unemployment rates of Roma and non-Roma by sex and age, 2011 (%) ...41 Figure 19: Jobless rates of Roma and non-Roma by sex in Central and Southeast Europe, 2011 (%) .............................................................42 Figure 20: Jobless rates by age in Central and Southeast Europe, 2011 (%) .....................43 Figure 21: Jobless rates by education in Central and Southeast Europe, 2011 (%) ........44 Figure 22: Informal employment rates among Roma and non-Roma in Central and Southeast Europe, 2011 (%) ..............................................................45 Figure 23: Inpatient stays, 2011 (%) .................................................................................................46 Figure 24: Access: health insurance coverage, 2011 (%) ..........................................................47 Figure 25: Access: affordability of buying medicines when prescribed, 2004-2011 (%) .. 48 Figure 26: Households without improved water source and sanitation (%), 2011 .......49 Figure 27: Roma households without access to... (%), 2011 ...................................................50 Figure 28: Insecure housing of households (%), 2011 ..............................................................51 Figure 29: Progress in Roma survey respondents’ access to basic infrastructure, 2004-2011 (in percentage points) ...............................................................................52

10

Figure 30: Material deprivation index (Roma and non-Roma, 2011)...................................53 Figure 31: Vulnerable groups and social exclusion ....................................................................57 Figure 32: Multidimensional poverty rate and its composition for Roma (non-severe and severe, bars left scale) and the value of MPI (right scale) (%) ......60 Figure 33: Multidimensional poverty rates (total) for Roma and non-Roma, 2004-2011 (%) .....................................................................................................................61 Figure 34: Contribution of individual dimensions to multidimensional poverty of Roma (2004-2011) ........................................................................................................63 Figure 35: Individual indicators’ contribution to multidimensional poverty reflected in the value of the weighted MPI .............................................................64 Figure 36: Correlation between multidimensional poverty and achieved level of education of adults (2011) (%) ..........................................68 Figure 37: Correlation between multidimensional poverty and achieved level of education of youth (2011) (%) ...........................................69 Figure 38: Gross enrolment rates in respective education levels, 2011 (%) ......................69 Figure 39: Average years spent in education by poverty status and age group, 2011 ..... 70 Figure 40: Unemployment rates by multidimensional poverty status and age group, 2011 (%) .................................................................................................71 Figure 41: Share of those who have never worked in the unemployed population in the respective age groups, 2011 (%) .......72 Figure 42: Different aspects of informality, 2011 (%).................................................................72 Figure 43: Share of respondents who said that stealing food is acceptable if it prevents a family from going hungry, 2011 (%) ..............................................74 Figure 44: Types of employment aspirations, 2011 (%) ............................................................74 Figure 45: Educational aspirations of multidimensionally poor parents (Roma and non-Roma) for their children (boys and girls), 2011 .......................76 Figure 46: Educational aspirations of multidimensionally non-poor parents (Roma and non-Roma) for their children (boys and girls), 2011 .......................77 Figure 47: Monetary and multidimensional poverty rates of Roma, 2004-2011 (expenditure-based) (%) ...........................................................................81 Figure 48: Overlaps between absolute, relative and multidimensional poverty estimates (%) ......................................................................................................82 Figure 49: Overlaps between material deprivation and multidimensional poverty estimates, 2011 (%)...........................................................................................84 Figure A: Average years spent in education by poverty status for the age 25-64 ..........92 Figure B: Share of the unemployed as a percentage of those in the labour force (15-64).............................................................................93 Figure C: Share of employed people who do not have a written contract (ages 15-64) .........................................................................................................................94 Figure D: Share of respondents who said that it is fully acceptable stealing food if a family goes hungry .........................................................................95

11

12

1

Introduction

This paper closes the cycle of analytical papers elaborated under the UNDP regional project “Decreasing Roma Vulnerability”, funded by the EU DG “Regional and urban policy” and the United Nations Development Programme. It builds on the findings of earlier publications in the working papers series, adding a new dimension to their sector-specific angle – that of human development. The paper goes beyond the standard (and still prevailing) approach conceptualizing poverty in terms of poverty rates. It presents the basic concepts and approaches to the definition of poverty; applies them to the Roma using data on the status of Roma households from surveys conducted by UNDP, the World Bank and the European Commission; and proposes a multidimensional measure that better reflects the multidimensional challenges Roma are facing. The term “poverty” has become synonymous with the term “Gypsy” or in the last two decades – “Roma” as these communities have come to be called. The change was an attempt to escape the pejorative connotations of “Gypsy”, which is firmly associated with an “underclass” but that change was not matched by dramatic decline of poverty – the Roma are similarly poor and marginalized as the Gypsies before. In response to these challenges reducing Roma poverty was put on the political agenda and formulated as an explicit commitment of the countries participating in the Decade of Roma Inclusion (2005-2015). The European Commission later joined in with its EU Framework of Roma Inclusion, adding broad political weight behind the attempts to lift Roma out of poverty. But what poverty are we talking about? What is “poverty”? Apparently an easy question, but its simplicity is misleading. This misleading simplicity often drives researchers, activists, and policy makers towards equally simplistic schemes for addressing the issue. The current analysis defines the Roma poverty in broader terms. Using the data from regional surveys among Roma communities at risk of marginalization and their nonRoma neighbours in Central and Eastern Europe, it goes beyond intuitive understandings of the issue. An adequate definition (as with an adequate diagnosis) is a precondition for an adequate intervention. This is particularly important today, in light of the European Commission’s new programming period, when the resources devoted to addressing Roma poverty will be larger by an order of magnitude—as, respectively, the potential for possible damage as a result of misguided policies.

13

Various approaches to poverty estimation There are various ways of defining and quantifying poverty. Most of these yield different results and thus often contributing more to the confusion around the issue than helping explain its root causes. The choice of method of measurement and setting poverty thresholds has obvious implications for the outcomes of the analysis and its policy implications. In principle, all poverty definitions can be classified according to two main characteristics: (i) the classification criteria; and (ii) the choice of the threshold (the numerical value reflecting the selected poverty criterion, below which the subject of analysis is considered “poor”). As regards to definitions, poverty estimates can be objective (defined by characteristics that are objectively measurable and not subject to an individual’s interpretations) or subjective (perception-based measures in which the assessment of the individual determines his or her poverty status). The first approach reflects some objective criteria (e.g., calories consumed, the costs of a “minimum consumption basket”, or disposable income). The second approach reflects the argument that objective criteria cannot capture poverty, and that only individuals themselves can say whether or not they are poor (one might live in a rich country with high income but be under pressure due to the presence of unaffordable consumption standards, and thus assess his/her status as “poor”).2 The data for both objective and subjective measures are usually derived from surveys. The similarity of the source shouldn’t however blur the difference between the two. In the first case, the status is determined by an external agent (an analyst assessing the data reported by the object of the poverty analysis). In the second case the status is determined by the person reporting his or her status. The number of dimensions of poverty being considered can also determine different approaches to poverty definition. A uni-dimensional poverty analysis (as the name suggests) takes a single aspect of poverty as a sufficient proxy—be it money (monetary approaches), natural units like calories of daily consumption, etc. (Townsend 1962: 215-220; Ravallion, Bidani 1994: 77-78; Lister 2004: 11-15). Such approaches are easily understood but usually too simplistic, reducing complex phenomena to a single variable. Multidimensional poverty measures are more comprehensive approaches to wellbeing, but their complexity comes at the cost of difficulties in aggregating the various indicators. Depending on how poverty thresholds are defined, one can distinguish between absolute and relative poverty. In both cases the thresholds are numeric cut-off values, with differences in how the value is determined. The absolute poverty thresholds are

2/ A sub-group of the “subjective” approaches is objective poverty estimates (using a numeric threshold) but applying a “subjective poverty line.” The latter is derived from a survey question “what income level do you personally consider to be absolutely minimal?” (Ravallion 1992: 33; Ravallion 2010: 12-19).

14

INTRODUCTION

derived from the concept of “needs” and imply some universal standard and some absolute minimum of unacceptable deprivation. A widely used measure for cross-country comparisons is the threshold of $2.15 per day in purchasing-power parity (PPP) terms, or the higher amount of PPP$4.30/day, below which one is considered severely poor or “just” poor.3 In the case of the relative poverty approach, the threshold determining whether a person is poor is set relative to the status of other members of the society. The median (of income or expenditures) is usually used as the reference point and the most commonly used threshold is 60% of the median (Garroway, de Laiglesia 2012: 1922). In other words, relative poverty captures a lower living standard in comparison to other people in that country, which does not necessarily imply a low standard of living or deprivation.4 In a response to the shortcomings of monetary estimates of poverty, a broad range of alternative approaches have been elaborated since the mid-1970s. One of the most prominent is associated with the human development concept, which attempts to reflect non-material (as well as material) aspects of human progress. While its roots reach back to Aristotle, the human development concept was operationalized for policymaking purposes only in 1990, when UNDP published its first human development report. The human development concept has evolved in the intervening two and a half decades, both in terms of defining what “human development” is and how to measure it (Alkire 2010: 8-11). The human development concept was largely inspired by Amartya Sen (Sen, 1987), who has argued that well-being comes from the capability to function in society. Thus, poverty arises when people lack key capabilities, and so have inadequate incomes or education, or poor health, or live in insecurity with low selfconfidence and a sense of powerlessness, or the absence of rights such as freedom of speech. All this limits people’s capabilities and ultimately reduces their freedoms (Sen 2000a: 87-110). However human development and the human development index (HDI/UNDP) are not measures of poverty or deprivation. HDI is an indicator of achievement, or progress (and not of deficits and poverty). In an attempt to address the different features of deprivation in the quality of life and arrive to an aggregate judgment on the extent of poverty, UNDP in 1997 introduced the Human Poverty Index (UNDP 1997: 17). The HPI was later “diversified” into the HPI-1 (for developing countries) and HPI-2 (for developed 3/ Purchasing-power parity takes into account not only a country’s exchange rate with the USD but also the cost of a similar (cross-country) basket of goods in order to determine a person’s actual purchasing power; it allows for a comparison of income from a relatively wealthy country with higher prices like the US with a country with lower income but also lower prices like Moldova for example. The UNDP MDGs and this paper use the International Comparison Program (ICP) initiated by the UN to construct PPP conversion factors (see http://siteresources.worldbank.org/ICPEXT/Resources/ICP_2011.html for more information). 4/ Absolute and relative poverty approaches can be synthesized if both are applied in a complementary manner through a “weakly relative poverty” in which the “elasticity of the poverty line to the mean is positive above some critical value, rises with higher mean consumption, but only reaches unity as mean consumption goes to infinity” (Ravallion 2010: 18-20).

15

ROMA EDUCATION IN COMPARATIVE PERSPECTIVE

countries), and eventually dropped in favour of the multidimensional poverty index. Multidimensional poverty encompasses a range of deprivations that a household may suffer. It was inspired by the need to identify a unified measurement of poverty for both developing and developed countries. The Oxford Poverty and Human Development Initiative was at the forefront of developing and promoting this methodology, which is now being broadly used for poverty monitoring beyond material wellbeing. Multidimensional poverty concept applies the capability approach in poverty monitoring, while also addressing a key conceptual drawback of the uni-dimensional approaches—namely, the loss of information on dimension-specific poverty measures (Alkire, Foster 2009: 8). The number of indicators and the specific indicators used depend on the purpose of the measure. The concept of social exclusion, which adds another dimension to poverty analysis, is usually included as an aspect of multidimensional poverty analysis. It has evolved together with the concept of social rights, rooted in the idea of the European welfare state. In discourses on citizenship, social rights and social justice, “being socially excluded” is not merely understood as a lack of access to goods, but as a lack of access to rights. Social exclusion may reduce human capabilities in one field, which might be responsible for deprivations in other fields of life, further fuelling the process of social exclusion. Amartya Sen refers to this as “capability failures”, and believes that social exclusion plays an instrumental role. Social exclusion is multi-dimensional, including economic, social and civic dimensions (Sen 2000b: 12-15). Deprivations in one dimension can reinforce deprivations in another, and these multiple deprivations can result in social exclusion. Table 1 summarizes the major approaches to poverty quantification and monitoring— and this summary is far from exhaustive. It suggests that there might be at least sixteen different answers to the question “What’s the poverty rate in a country X or for group Y?”—all of which could be correct in their own right. However, when communicated without an explanation of the methodology applied, without the relevant metadata available or detached from the specific context (which is most often the case) poverty estimates can be used in a highly manipulative manner.

Institutional clustering of poverty monitoring methods While the World Bank is traditionally seen as an advocate of absolute poverty measures, in reality the picture is more nuanced. According to the World Bank, “poverty is pronounced deprivation in wellbeing” (World Bank. 2000: 15). This definition builds on the Bank’s earlier approach defining poverty more narrowly, as “the inability to attain a minimal standard of living” (World bank 1990: 26). Both definitions however raise two questions about what is meant by well-being (or living standard) and what is the reference point against which to measure deprivation (or inability). Two approaches are typically used. One approach is to think of well-being as the command over com-

16

INTRODUCTION

Table 1: Different approaches to poverty

Income based

National thresholds specif- 1. Nationally specific ic for individual countries, income-based poverty in the national currency rates Absolute poverty lines

Monetary

Absolute poverty lines

Relative poverty lines

Multidimensional

60% of the median income

2. Severely poor with income below 2.15 PPP$ 3. “Just poor” with income below 4.30 PPP$ 4. Severely poor (below 40% of the median income) 5. “At risk of poverty“ (below 60% of the median income)

National thresholds specif- 6. Nationally specific ic for individual countries, expenditure-based in national currency poverty rates Expenditure based

Uni-dimensional

Relative poverty lines

Internationally comparable thresholds

Internationally comparable thresholds

Share of the median expenditure

7. Severely poor with expenditures below 2.15 PPP$ 8. “Just poor” with expenditures below 4.30 PPP$ 9. Severely poor (below 40% of the median expenditures) 10. “At risk of poverty“ (below 60% of the median expenditures)

Food energy intake (FEI)

11. Nationally specific FEIbased poverty rates (varies by climate conditions, rural/urban distribution, type of occupation etc.)

Basic needs and human development

12. HPI1 and HPI2

Social exclusion

13. Social exclusion index

Multidimensional poverty estimates – internationally comparable (following the methodology developed by OPHI and used for international comparisons and in the Global HDRs published by UNDP)

14. Multidimensional poverty index

Nationally specific, following the methodology developed by OPHI

15. Severely poor 16. “Just” poor

17

modities in general, so that people are better off if they have a greater command over resources. The main focus is on whether households or individuals have enough resources to meet their needs. Typically, poverty is then measured by comparing individuals’ income or consumption with some defined threshold below which they are considered to be poor. This is the most conventional view—poverty is seen largely in monetary terms—and is the starting point for most analyses of poverty. A second approach to well-being (and hence poverty) is to ask whether people are able to obtain a specific type of consumption good: Do they have enough food? Or shelter? Or health care? Or education? In this view the analyst goes beyond more traditional monetary measures of poverty: nutritional poverty might be measured by examining whether children are stunted or wasted; educational poverty might be measured by asking whether people are literate or how much formal schooling they have received. For international comparisons, the World Bank uses PPP$1.25/day and PPP$2.30/day poverty thresholds.5 These thresholds were introduced in 1990, originally as $275 and $370 per person a year in constant 1985 PPP$ prices (World Bank 1990: 27). Unlike the World Bank, the European Union emphasizes relative poverty. The at-riskof-poverty rate is the share of people with an equivalent disposable income (after social transfer) below the at-risk-of-poverty threshold, which is set at 60% of the national median equivalent disposable income after social transfers. This indicator does not measure wealth or poverty, but rather low incomes in comparison to other residents in that country (which does not necessarily imply a low standard of living).6 The Oxford Poverty and Human Development Initiative (OPHI) is among the organizations pioneering multidimensional poverty approaches. It defines poverty in terms of multiple and overlapping deprivations that a household may suffer.7 The number and type of indicators used depend on the purpose of the measure. These include national poverty measures that reflect changes over time, targeting of services or cash transfers, and monitoring and evaluation. At a glance, multidimensional measures present an integrated view of the situation. The multidimensional poverty index (MPI) measures overlapping deprivations at the household level across the same three dimensions as the Human Development Index (living standards, health, and education) and shows the average number of poor people and deprivations with which poor households contend (Alkire, Foster 2011: 5-9). One deprivation alone may not constitute poverty; poverty as per the MPI requires that a household be deprived across multiple indicators at the same time. A person is multidimensionally poor if he or she is deprived in at least one third of the weighted indicators. 5/ See http://data.worldbank.org/indicator/SI.POV.DDAY, http://data.worldbank.org/indicator/ SI.POV.2DAY, and http://iresearch.worldbank.org/PovcalNet/index.htm 6/ http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Glossary: At-risk-of-poverty_rate http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Income_distribution_statistics 7/ http://www.ophi.org.uk/research/multidimensional-poverty/

18

INTRODUCTION

The UNDP adopts a mix of approaches. Although it is not defined in “calorie intake” terms, MDG 1 implicitly calls for a consumption-based approach. The HDI is a composite index and crude deprivation proxy measure, which has undergone a number of modifications since its introduction in 1990. Since 2009, the UNDP includes MPI in the global human development indicators monitored in its Human Development Reports. Due to the highly decentralized nature of the organization, individual country offices conduct their national-level analyses and NHDRs following diverse approaches (usually reflecting national specifics and approaches).

The conceptual framework behind this analysis The above brief overview shows that the methodological choices matter not just for the evidence their application produces, but also (perhaps most of all) for their policy implications. Those choices need to reflect the nature and the specifics of the phenomenon being researched (e.g., Roma poverty and exclusion) and the purpose of the analysis. Multidimensionality is a major characteristic of Roma poverty. Roma poverty is not just a lack of financial resources, unemployment, substandard housing, or poor access to social services—it’s about the combination and interactions of all these factors, which are both outcomes of past exclusion and determinants of future deprivations—reinforcing the vicious circle of poverty. This multidimensionality requires similarly multidimensional conceptual approach to integrate poverty reduction and fundamental rights agendas. Unemployment, social exclusion, and marginalization are interlinked with (and mutually reinforce) discrimination, anti-Gypsyism, limited access to justice and segregation. The human development paradigm can serve as such a multi-dimensional approach. In the same spirit, this analysis does not just wish to show that Roma communities (and particularly those vulnerable to marginalization) are among the poorest in Europe. Its purpose is to look into the mutual relationships among variety of socioeconomic characteristics in order to build a better picture of the determinants of poverty. Once those determinants are better understood, it would be possible to suggest practical interventions to address them—leading to a decline in poverty. Sketching the true picture of poverty, answering the question “Who are the poor?” is a necessary precondition for adequate poverty reduction interventions. For that purpose, detailed socioeconomic profiles of the “MPI poor” and “MPI non-poor” have been constructed and analysed. The paper then develops an econometric analysis to determine the contribution of each individual dimension to the risk of poverty.

The sources and the constraints of this research “There are no quantitative data on the situation of Roma households” is a popularly– although incorrect--view. There is a great deal of data. What is missing is comparability

19

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

and methodological consistency. Studies rarely use the same (or similar) methodology and tools. Data from different “Roma” surveys are therefore rarely comparable, both vis-à-vis one another and relative to the results produced by such standard statistical tools as household budget or labour force surveys. The data from the 2011 regional Roma survey, which was completed by UNDP in cooperation with the World Bank and the European Commission (EC), and in coordination with the European Union’s Agency for Fundamental Rights (FRA), and which are methodologically comparable to the data produced by UNDP’s 2004 regional Roma survey are an effort to fill this gap. The two surveys, conducted in 2011 by UNDP/WB/EC and FRA, represent the largest integrated household-level data collection efforts on the status of the Roma to date. The combined UNDP/WB/EC regional Roma survey and the FRA Roma pilot survey (both from 2011) has data produced by 20,018 Roma households (with 87,717 household members living in these households) and 9,782 non-Roma households living nearby (28,214 household members), covering 18 European countries. The 2011 surveys followed the conceptual approach tested in 2004 by UNDP in its first comprehensive regional survey of Roma at risk of marginalization and non-Roma living in close proximity (UNDP 2006).8 This analysis is based on data from these two surveys of Roma households at risk of marginalization and non-Roma living in close proximity. There are several reasons for using these two datasets. First, the data are comparable for all countries of Central and Eastern Europe, which allows for an analysis to be made over time for these countries.9 Second, this analysis refers to a number of in-depth thematic regional studies, which are based on the same data set and provide additional insights into the sector-specific dimensions.10 Third, the interval of seven-years is more appropriate for capturing any changes after the launch of the “Decade of Roma Inclusion” in 2005. This was the period in which the issue of full and equal participation of Roma in their societies was put firmly on national policy agendas. In some ways these efforts are paying off, although the results are far from initial expectations. Fourth, the UNDP survey complements the data on the status of households with information about the behavioural patterns and values of the respondents, which can in turn be linked to their socio-economic status. Finally, the survey was administered on two sample groups: Roma at risk of marginalization, and non-Roma living in close proximity (and therefore sharing the socioeconomic characteristics of the areas inhabited by the Roma surveyed). Thus, the data 8/ The 2004 survey was prepared and implemented by UNDP. The 2011 survey was implemented jointly by UNDP, the World Bank and the European Commission, Directorate General “Regional and Urban Policy”. 9/ The comparability of the two surveys is based on similar methodologies applied in both time periods and not the ideal of a similar panel of households. To the authors’ knowledge this has yet to be accomplished due to the sensitivity of Roma households and European policy makers in retaining personal addresses of ethnic groups for longitudinal panels. 10/ The papers are accessible from the UNDP web-site, http://www.undp.org/content/rbec/en/ home/ourwork/povertyreduction/roma-in-central-and-southeast-europe/

20

INTRODUCTION

allow for multiple levels of comparability: between the two groups, and on core socioeconomic indicators—also between these groups and the national averages, and over time.11 For issues related to “discrimination,” the analysis draws on data from the FRA survey conducted in parallel to that of UNDP. Both the FRA and UNDP surveys follow the methodological framework for defining the universe of study (the population to be studied) first proposed by the UNDP team in 2004: a combination of external (territorially determined) criteria (such as segregated Roma neighbourhoods), and self-identification (self-determination as “Roma” at the beginning of the interview).12 This approach to defining the universe of study was also applied in the FRA study in 2011. Thus, all three surveys examine not an abstract “Roma” population, but Roma at risk of marginalization due to their segregated status. This makes their results particularly relevant in terms of integration policies in their respective countries.

11/ The survey followed the format of an integrated household survey; a number of questions in the questionnaire are identical to similar nationally representative surveys. For more information about the study and its methodology see Ivanov et.al (2012). 12/ See UNDP (2006), http://europeandcis.undp.org/poverty/show/A3C29ADB-F203-1EE9BB0A277C80C5F9F2; FRA (2014). Roma Pilot Survey – Technical report: methodology, sampling and fieldwork. http://fra.europa.eu/en/publication/2014/roma-pilot-survey-technical-reportmethodology-sampling-an

21

2

LOOKING AT POVERTY FROM A MONETARY PERSPECTIVE

The data presented in Figures 1-1213 provide a comprehensive picture of the magnitude and depth of poverty of Roma at risk of marginalization and their non-Roma neighbours. The analysis presents both monetary (absolute and relative) and multidimensional poverty profiles. The country profiles provide an aggregatFigure 1: Absolute monetary poverty (expenditure based) of Roma and nonRoma, 2004 and 201114 (%)

13/ The monetary poverty indicators are based on the household income and expenditures data from the 2004 and 2011 Regional Roma surveys. The income was estimated based on the question “I will ask you now about the income (or transfers) your household actually received (net income) during the last month. Please tell me, what sum, approximately, was made by each of these kinds of incomes (or transfers)?” for 2004 and “Please tell me, what were the main sources of these incomes of your household (estimate roughly)? What were the approximate MONTHLY amounts?” for 2011. The expenditure was estimated based on the question “How much money did your household spend last month in total?” for 2004, “And how much money did your household spend last month in total?” for 2011. For estimating the equivalent income and expenditure, the following weights were assigned based on the modified OECD equivalence scale – it gives a weight of 1.0 to the first adult, 0.5 to any other household member aged 14 and over and 0.3 to each child aged less than 14.

22

ed picture of poverty status in 2004 and 2011 as captured by the two rounds of the survey. Most of the data generated by the surveys is comparable and allows the tracking of progress between 2004 and 2011 along individual dimensions. The country profiles point to significant differences across countries, attributable to a number of factors— especially overall levels of socioeconomic development, and comprehensiveness and quality of Roma inclusion policies.

Absolute poverty profiles The two approaches to calculating monetary poverty (by income and by expenditures) yield a generally consistent picture. Income estimates usually produce higher poverty rates (people tend to underreport income)—and this is also the case in both datasets, of the Roma and non-Roma. In all countries with the exception of Bosnia and Herzegovina both approaches show similar trends of declining poverty. In the case of Bosnia and Herzegovina, however, income-based estimates show an increase in income-based poverty among Roma by 16 percentage points and a decline in expenditure-based poverty by 7 percentage points (Figures 1 and 2). Figure 2: Absolute monetary poverty (income based) of Roma and non-Roma, 2004 and 2011 (%)

14/ For visual clarity, the following abbreviations were used in the graphs: AL (Albania), BA (Bosnia and Herzegovina), BG (Bulgaria), H (Hungary), HR (Republic of Croatia), CZ (Czech Republic), MD (Moldova), ME (Montenegro), MK (Former Yugoslav Republic of Macedonia), RO (Romania), Î

23

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Expenditure-based estimates are accepted as generally more reliable for vulnerable groups. These data suggest that significant progress has been made in reducing the magnitude of Roma poverty during 2004-2011. The share of people living on less than PPP4.30/day (measured by expenditures) declined in all countries (Figure 2, p. 23). The most substantial decline in Roma poverty occurred in Albania, where the distance between the Roma and their non-Roma neighbours was reduced the most dramatically. The most modest progress was recorded in Montenegro and FYR Macedonia. Progress in poverty reduction for non-Roma respondents was not so apparent. In some countries the incidence of poverty for this group increased (in Bosnia and Herzegovina and Montenegro) or stagnated (in Croatia and FYR Macedonia). This lack of progress or deterioration may be due to a variety of reasons that cannot be tracked with the available data. A more thorough analysis of the local context is needed for that purpose. It is a worrying sign in any case and unless both Roma and non-Roma benefit from poverty reduction initiatives, popular support for Roma-targeted measures may weaken. The structure of poverty (extreme or otherwise) provides additional insight into the dynamics of monetary poverty between 2004 and 2011. Here too, the two approaches (income and expenditure based) yield somewhat different results. Based on expenditures (Figure 3), the countries that registered the greatest decline in poverty also experienced a substantial decline in absolute poverty (and respectively a decrease in the poverty gap15). These are Albania, Serbia, and Bulgaria. One might say that in these countries poverty reduction measures captured the entire spectrum of the poor. In the other group of countries (Bosnia and Herzegovina, Croatia, FYR Macedonia, and Montenegro), reduction in poverty was achieved primarily for the population segment below the PPP$4.30/day but above the PPP$2.15/day thresholds, minimally reducing extreme poverty rates (if at all). Poverty reduction policies in those countries may have exhausted this low hanging fruit (it is always easier to get those closest to the poverty line above it). If this is the case indeed, further reductions in Roma poverty may be unlikely if policies are not explicitly targeted at the most vulnerable Roma households. Based on income (Figure 4), extreme poverty in the second group of countries has even increased (particularly in Bosnia and Herzegovina, where the rise in poverty incidence was matched by an increase of 10 percentage points in the poverty gap).

Relative poverty profiles Relative poverty reveals yet another aspect of the poverty picture. The data presented Î RS (Republic of Serbia), and SK (Slovakia). The abbreviations are following the country codes used by EUROSTAT, http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/ Glossary:Country_codes. In all graphs depicting individual countries, the countries are sorted by alphabetical order. 15/ The poverty gap is a measure of the shortfall of those in poverty from the poverty line expressed as a percentage of the poverty line and counting the nonpoor as having zero shortfall. A reduction in the poverty gap means that the population is coming closer to the poverty line.

24

LOOKING AT POVERTY FROM A MONETARY PERSPECTIVE

Figure 3: Monetary poverty and severe poverty rates (expenditure based) and poverty gap for Roma, 2004 and 2011 (%)

Figure 4: Monetary poverty and severe poverty rates (income based) and poverty gap for Roma, 2004 and 2011

25

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 5: Different dimensions of inequality: relative poverty rates for Roma, their non-Roma neighbours and the country averages (selected countries), 2011 (%)

in Figure 5 reveal deep disparities between Roma, their non-Roma neighbours and the general populations in all countries. Compared to national averages, the relative poverty rates for both groups are far higher. The relative poverty rates were on average between 3 and 4 times higher for the Roma and twice higher for their non-Roma neighbours than the national “at-risk-of-poverty rate by poverty” for the respective countries (as reported by Eurostat). Of particular interest is that the difference in poverty rates between Roma across countries is not as great as the difference between non-Roma. This suggests that Roma at risk of marginalization are similarly poor, regardless of the country they live in. Apart from inequalities between groups, significant horizontal inequalities exist (within

26

LOOKING AT POVERTY FROM A MONETARY PERSPECTIVE

groups). Figure 6, visualises the Gini coefficient for both groups and the country averages. It also underscores the importance of horizontal inequality, in this case among Roma and non-Roma. Concerning intra-group inequality, the picture is mixed. These inequalities are higher among Roma in less developed countries (Bulgaria, Romania, Moldova). In the rest the value of the Gini coefficient is equal or close to equal for both groups.16 Figure 6: Magnitude of horizontal inequality: Gini coefficients for Roma, their non-Roma neighbours and the national averages (selected countries17), 2011 (%)

16/ The Gini coefficient measures inequalities within a given population; here, it measures income inequalities of Roma and Non-Roma separately. The higher the coefficient, the greater the inequality within each population. 17/ The data for Bulgaria, Czech Republic, Croatia, Hungary, Romania and Slovakia are from the Eurostat databases and apply for the year 2011. The data for Macedonia, Montenegro, Serbia, Albania, Bosnia and Herzegovina and Moldova are from the World Factbook and 2011 or the closest year available was used (2011 in all cases but FYR Macedonia, 2010 in the case of Montenegro and Moldova, 2007 for Bosnia and Herzegovina and 2008 for Albania and Serbia was used.

27

ROMA EDUCATION IN COMPARATIVE PERSPECTIVE

Figure 7: Structure of Roma household income, 201118 (%)

Figure 8: Structure of non-Roma household income, 2011

18/ In the Czech Republic, Slovakia, Hungary, Croatia, Romania, Montenegro, Bulgaria, Serbia, Albania, Bosnia and Herzegovina, FYR Macedonia, and Moldova.

28

LOOKING AT POVERTY FROM A MONETARY PERSPECTIVE

Income and expenditures How much money a household has is important. Equally important are the sources of the income and what income is spent on. Earnings from employment constitute 24% and 34% of the total household income of Roma and non-Roma respectively (Figures 7 and 8). When income from labour activities other than employment is added however, the total share of labour income is very close: 36% for the Roma and 38% for the nonRoma. This suggests that the stereotype of Roma not working is exaggerated, to say the least. The large differences are found in income from pensions (41% of the income of nonRoma and only 18% of the income of Roma), child benefits (respectively 5% and 13%) and social assistance (respectively 9% and 21%). Unemployment benefits constitute a small share of income—3% in the case of non-Roma and 6% of Roma. In other words, both groups are equally dependent on social transfers (unemployment benefits, pensions, social assistance and child allowance).19 For both the non-Roma and Roma they constitute 58% of total income. The difference is in their structure. The transfers for non-Roma are dominated by pensions and for the Roma – by child allowances and general social assistance. In the long-run one might expect the share of the pensions in household income to decline further given the impact of the systemic transformation and the collapse of the socialist centrally-planned economies or welfare state retreatment and privatization of pensions systems. The (few) Roma pensioners today are the ones who earned their retirement in the previous regimes; the potential future pensioners (today’s youth) is mostly unemployed (as shown further in this paper). Thus the message “Roma are abusing the social protection system” is also hard to support. Both groups depend on the state for their survival; put simply their demographic profiles are different. This also determines the different nature of the social transfers. Those for the non-Roma are devoted primarily to providing a decent (if possible) living for a generation that has completed its active employment activity, thus these transfers are oriented in that regard towards the past. In the case of the Roma, the transfers are targeted more explicitly at the younger generation and should help it to develop its skills and talent. In that regard it is (in theory at least) an investment from which the society will benefit in the future. This is the logic behind “Roma inclusion as smart economics” (De Laat and Bodewig, 2011) putting forward economic arguments in favour of Roma inclusion (analysed in-depth below). Whether this theory is implemented in practice is a different question. 19/ Theoretically, pensions can be considered a type of social insurance (employees contribute during the active professional life to receive secured income after retirement). But these are funded pension schemes in which the contributions are invested in a fund towards meeting the benefits. In most CEE countries the pension schemes are mixed with the first pillar being unfunded (traditional PAYG system in which the benefits of the retired are paid from the contributions of the employed). The deficits in the social insurance institutes managing the first pillars are covered from the state budgets. In addition, the more vulnerable a person, the higher the relative weight of social or disability pensions, which bear the characteristics of social transfers. This is why in the context of the current research pensions are considered social transfers.

29

Figure 9: Structure of Roma household expenditures, 2011

Figure 10: Structure of non-Roma household expenditures, 2011

30

LOOKING AT POVERTY FROM A MONETARY PERSPECTIVE

The structure of household expenditures is very similar for both groups (Figures 9 and 10). In both cases they are dominated by spending on food and everyday household goods and services, which constitute 56% of the total expenditures of Roma and 49% of non-Roma households. The second largest portion of the monthly budget goes to housing and utilities, and constitutes 17% of Roma and 21% of non-Roma household expenditures. The fact that non-Roma and non-poor Roma spend a larger percentage of total expenditures than other Roma is likely due to the former groups’ larger living spaces and more expensive heating and cooking options. Surprisingly, both groups spend more money on alcohol and cigarettes (9% for Roma and 6% for non-Roma) than on education, transportation, and clothes, including shoes. Roma households, who report smaller amounts of total expenditures, could benefit from reallocating these expenditures toward the accumulation of human and physical capital. Figure 11: Structure of Roma and Non-Roma household income by poverty status, 2011

31

ROMA EDUCATION IN COMPARATIVE PERSPECTIVE

Figure 12: Structure of Roma and non-Roma household expenditures by poverty status, 2011

Examining these data by absolute poverty status20 shows that the structure of income among non-Roma is primarily differentiated by pension income. However, severely poor non-Roma look similar to the Roma with small pensions and larger income shares coming from social assistance. The income structure for severely poor Roma looks even more like that of the severely poor non-Roma, which suggests that the poverty reduction challenge may entail involving similar policy instruments in the case of both groups. Predictably, earnings related to employment are associated with lower poverty rates among both the Roma and non-Roma. Interestingly, the shares of income derived from “other labour activities” are equally high among all poverty groups among the Roma 20/ Respondents living below PPP$2.15/day are classified here as “severely poor”; those living between PPP$2.15/day and PPP$4.30/day are classified as “poor”; while those living above PPP$4.30/day are classified here as “non-poor”.

32

LOOKING AT POVERTY FROM A MONETARY PERSPECTIVE

and the severely poor non-Roma suggesting that people from these four groups are equally dependent on informal employment or self-subsistence activities. The structure of expenditure does not change significantly according to poverty status (Figures 11 and 12). The largest contrast is the smaller share of expenditures on food by the Roma non-poor. This makes sense and allows for more spending on housing and utilities. This difference however, is only a few percentage points. Again, alcohol and cigarettes have quite a large share—even larger than the share for medicine and medical expenses. But the fact that they have a similar share across poverty status indicates that the poor, and the Roma, are no different than others in their consumption of these items.

33

3

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Monetary approaches to defining poverty make sense under the assumption that financial resources can be (and are) translated into goods and services that make life more complete and meaningful. According to the human development paradigm, money is purely a means for achieving a meaningful life. Similarly, the capability approach looks at poverty as a deprivation in functionings (the “beings and doings” according to Amartya Sen, the states and activities constitutive of a person’s being). Financial resources may be used as a proxy for capabilities, but these capabilities do not necessarily transform into functionings because consumption patterns do matter. A household may be financially not poor, but might spend its income on items that do not expand its members’ functionings (like the expenditure on alcohol and cigarettes in the example above). In addition, money-metric approaches do not include public goods consumption, particularly on social services like health and education, or public infrastructure (Thorbecke 2005: 5). In addition, the standard of living does not depend exclusively on the availability of financial resources. Goods and services can be produced and traded in kind or simply misappropriated. For example, when a person travels on public transportation without a ticket, s/he consumes a transportation service that is not recorded in the statistics of household expenditures. This is why it is reasonable to look at the real picture of poverty not through the lens of the “universal equivalent” but directly, by analysing the real living standards of individuals and households.

Educationally poor21 The UNDP/World Bank/European Commission 2011 regional Roma survey data outline three major problem areas when comparing the educational participation of Roma and non-Roma: access to education, school completion and early school leaving. Gaps in these three areas mutually reinforce one another—low pre-school enrolment rates severely reduce the chances of school completion and increase the probability of dropping out of school early. Access to early childhood education and care appears to be a major challenge for Roma—particularly as non-attendance of pre-school clearly limits professional and educational opportunities later in life. Figure 13, presents data on the shares of Roma and 21/ This section is based on the findings in Brüggemann, 2012. The Figures visualise the data derived from the household grid registering the educational status of each household member.

34

Figure 13: Pre-school attendance (%) Shares of Roma and non-Roma children aged 3 to 6 who attended pre-school, kindergarten, or nursery (2011)

Sources: UNDP/World Bank/EC regional Roma survey 2011; UNICEF TransMONEE database for net enrolment rates. Notes: 1) Children who have already been enrolled in primary school were not considered in the calculation. 2) For the Czech Republic, Slovakia, FYR Macedonia, and Montenegro, the age group is 3 to 5 because the TransMONEE reference data refer to this age group. 3) For Slovakia and Albania, net enrolment ratios are based on data from the 2009/2010 school year. 4) Slightly different results regarding pre-school attendance are presented by FRA and UNDP (2012, p. 13). These differences stem from the fact that the FRA and UNDP results are based on use of a pooled dataset (combining UNDP and FRA survey data), and because a different age cohort has been used for the calculation. Nursery attendance is not included in the pooled dataset.

non-Roma aged 3 to 6 (or 3 to 5) who attended pre-school facilities (including nursery, kindergarten and mandatory pre-school classes a year before primary school) in 2011, as well as national average net enrolment rates. Roma were much less likely to attend pre-school than non-Roma living in close proximity. Differences in pre-school atten-

35

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 14: Roma with at least lower secondary education (%) Shares of Roma aged 17 to 23 who completed at least lower secondary education (ISCED 2) in 2004 and 2011

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011. Note: The UNDP vulnerable groups survey 2004 did not cover Slovakia or Moldova.

dance rates between Roma and non-Roma were statistically significant in all countries surveyed, except for Bosnia and Herzegovina, FYR Macedonia, and Montenegro (countries with the lowest overall pre-school attendance rates). The association between group affiliation and pre-school attendance was strongest in Moldova and the Czech Republic: in Moldova, 23% of Roma attended pre-school, compared to 82% for nonRoma. In the Czech Republic, 28% of Roma attended pre-school, compared to 65% for non-Roma. No significant gender differences were found (Brüggemann 2012: 31). School completion is another challenge directly associated with poverty. Figure 14 shows the shares of Roma aged 17 to 23 who completed at least lower secondary education in 2004 and 2011 (lower secondary education in the region refers to 4-5 years of schooling following primary education). Many Roma did not complete lower secondary education in 2011. Although more than 80% of Roma completed lower secondary education in the Czech Republic, Slovakia, and Hungary, 2009 national labour force survey data from Hungary and Slovakia indicated that only 1% of the overall population (aged

36

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 15: Early leavers from education and training (%) Shares of Roma and non-Roma living in close proximity aged 18 to 22 who are not attending school and have not completed education higher than lower secondary (ISCED 2) or short-term upper secondary (ISCED 3c)

Sources: UNDP/World Bank/EC regional Roma survey 2011; Eurostat (2013) for national averages in 2011. Note: National averages were not available for Albania, Bosnia and Herzegovina, Moldova, Montenegro and Serbia.

25 to 64) had not completed lower secondary education (OECD 2011, p. 38).22 By this standard, the shares of young Roma that did not achieve this education level (20% in Slovakia, 13% in Hungary) exceed the national averages. Educational attainment rates for Roma in the Southeast European countries were below those in Central Europe. Secondary educational attainment rates for Roma were below 50% in Romania, Bosnia and Herzegovina, Croatia, Montenegro, Albania, and Moldova (Brüggemann 2012: 22). Dropping out before completing compulsory schooling is another aspect of Roma educational poverty. In many countries, Roma are more likely to have completed shortterm upper secondary education and thus fall into the “early leavers from education and 22/ Despite differences of methodologies, the data from the regional survey can be compared to provide an idea of the magnitude of the existing gaps between the status of Roma and the national averages.

37

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 16: Ethnically segregated schools (%) Share of Roma and non-Roma living in close proximity aged 7 to 15 who attend regular schools with majority Roma student body

Source: UNDP/World Bank/EC regional Roma survey 2011.

training” category.23 As shown in Figure 15 (p. 37), most Roma (more than 80% in Hungary, Bulgaria, Bosnia and Herzegovina, Croatia, Macedonia, and Serbia for 2011) fall into the “early leavers” category. In the Czech Republic, Slovakia, Romania, Montenegro, Albania, and Moldova, this share rose to above 90%. The differences between Roma and non-Roma were statistically significant in all countries. The association between group affiliation and early school leaving was weakest in Hungary and strongest in Montenegro, where 94% of Roma were early school leavers, compared to only 29% for non-Roma living in close proximity. No statistically significant gender differences were found anywhere except Croatia, where the share of Roma women who leave school early was 10 percentage points higher than the share of the Roma men doing so. (Brüggemann 2012: 49) Educational segregation is another dimension of the educational poverty that Roma 23/ Early leavers from education and training refer to persons between the ages of 18 and 24 that are not enrolled in school or training, and have not attained at least an upper secondary education (ISCED 3) - Eurostat 2011, p. 203.

38

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

experience, and is both an outcome and a driver of Roma exclusion. Data suggest that in Slovakia, Hungary, Bulgaria, Romania, FYR Macedonia, Montenegro, Albania, and Moldova, the shares of Roma attending segregated schools was statistically significantly higher than the shares of non-Roma living in close proximity. In Hungary, Romania, FYR Macedonia, and Moldova, more than 20% of Roma students attended such schools (Figure 16). In Slovakia and Bulgaria, this share exceeded 30%. The association between group affiliation and segregated school attendance was strongest in Slovakia: 34% of Roma attended schools with a predominantly Roma student body, compared to 5% for non-Roma (Brüggemann 2012: 64). Apart from segregated schooling, Roma children often end up in “special” schools or classes that are ostensibly for children with disabilities, or other special needs. This special schooling constitutes per se segregated education because activities in these facilities are separated and different from those associated with regular education. They also offer reduced curricula and rarely enable their students to enter the regular school system or the labour market. According to the 2011 survey data, the shares of Roma aged 7 to 15 attending special schools (not including special classes) exceeded 5% in Hungary, Serbia, and Croatia, and 10% in the Czech Republic and Slovakia.24 A statistically significant decline in the shares of Roma attending these schools during 2004-2011 was noted in the Czech Republic—from 25% to 17%. However, a statistically significant increase in the shares of Roma attending these schools during this time was noted in Croatia—from 2% to 7%. No significant gender differences were found (Brüggemann 2012: 67). A particularly severe form of educational segregation occurs when Roma children make up the majority of the student body in special schools. The results of the survey strongly suggest that many special schools in the Czech Republic, Slovakia, Hungary, and Serbia are ethnically segregated. In all countries except Croatia, the share of Roma attending ethnically segregated special schools is higher than the share of Roma attending ethnically segregated regular schools. In the Czech Republic and Slovakia, over 60% of all Roma who attend special schools are subject to twofold educational segregation: they attend special schools with a predominantly Roma student body.25

Employment poverty26 Unsurprisingly, throughout Central and Southeast Europe, the Roma face higher unemployment rates than the non-Roma populations living in their close proximity. Roma are more likely to be unemployed than their non-Roma counterparts in all countries 24/ A separate UNDP household survey conducted in Slovakia 2010 based on a different sampling methodology found that 16% of Roma aged 7 to 15 attended special schools; another 4% were in special classes (Brüggemann and Škobla 2012, p. 2). 25/ In-depth studies at a country level provide an important and more detailed picture of this phenomenon – and it is even more worrisome compared to the survey data. For example, in the Serbia special schools report based on data collected 26/ This section is based on the findings in O’Higgins, 2012. The figures in this section are based on the individual responses in the employment module of the survey.

39

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 17: Unemployment rates for Roma and non-Roma in Central and Southeast Europe, 2011 (%)

Source: Roma and non-Roma percentages calculated from UNDP/WB/EC regional Roma survey 2011; National averages are drawn from Eurostat (Czech Republic, Slovak Republic, Hungary, Bulgaria, Romania and Croatia; epp.eurostat.ec.europa.eu) and ILO-KILM (Bosnia and Herzegovina, FYR Macedonia, Serbia, Montenegro, Albania and Moldova; www.ilo.org) databases. Notes: 1) The unemployed are defined on the basis of the standard ILO criteria; that is, as those who are a) without work, b) willing and able to work, and, c) actively seeking work; 2) The unemployment rate is the number of unemployed expressed as a percentage of the labour force for those within working age (15-64). 3) National averages are the annual average for 2011 except for Bosnia and Herzegovina and FYR Macedonia (2010), Serbia, Montenegro, and Albania (2009). 4) The “national” averages for Montenegro and Serbia are both the average for the two countries taken together.

40

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 18: Unemployment rates of Roma and non-Roma by sex and age, 2011 (%)

and for both men and women. Unemployment rates in relation to the national averages of both Roma and non-Roma populations living in close proximity vary across countries (O’Higgins 2012: 15, 16). This in part reflects the geographical distribution of vulnerable Roma communities which, for example, in Slovakia and Hungary, are concentrated in more impoverished parts of the country. Thus, both Roma and proximate non-Roma unemployment rates based on the UNDP/WB/EC regional Roma survey are significantly higher than the national averages (Figure 17). Examination of unemployment by age adds an important dimension to the issue. Data summarized in Figure 18 suggest that important gaps exist, both by ethnicity and by sex, and these gaps differ for different age groups. Non-Roma men and women have similar unemployment rates for most age groups, except for those aged between 2534, where the unemployment rate for men is more than double the rate for women. Also, in this age group the gap between Roma and non-Roma women is highest (almost three times). Interestingly, the gender gap in unemployment is more pronounced than the ethnic gap.

41

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 19: Jobless rates of Roma and non-Roma by sex in Central and Southeast Europe, 2011 (%)

Source: calculated from the UNDP/WB/EC regional Roma survey 2011. Notes: 1) the jobless rate and ratio are calculated on the working age (15-64) population 2) The jobless rate is defined as the proportion of the gender/ethnic specific population which is neither in education nor employment.

42

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 20: Jobless rates by age in Central and Southeast Europe, 2011 (%)

Source: calculated from the UNDP/WB/EC regional Roma survey 2011.

However, the ILO-defined unemployment rate is a problematic indicator for assessing the labour market situation of Roma. The ILO applies a restricted definition of the labour market and excludes all those who drop out of the labour market—thus understating the size of labour market problem.27 The jobless rate (the ratio of those not in employment or education to the relevant population) may be a more informative indicator to unemployment rates. Jobless rates are also much higher amongst Roma than their non-Roma neighbours (Figure 19). The cross-country differences are also somewhat attenuated particularly at the extremes. A Roma in the Czech Republic is “only” around three times as likely to be jobless as a non-Roma (O’Higgins 2012: 18-21).28 Similarly to unemployment, the analysis of joblessness by age reveals significant gaps by different groups, defined both by age and ethnicity. Although its prevalence is greatest amongst older workers, the gap between Roma and non-Roma is smallest for the aged 55-64 11 percentage points). The joblessness declines for younger cohorts but the difference between youth and old-age is much smaller among the Roma than among their non-Roma neighbours (Figure 20). In the case of non-Roma these are also higher than the national averages, although to a different degree. Eurostat data suggests that NEET rates for people aged 16-24 (those not in employment and not in education or training) in 2010 were 22% in Bulgaria, 8% in the Czech Republic, 13% in Hungary, 17% in Romania and 14% in Slovakia. Data for other countries is fragmentary. 27/ This has led the World Bank (2006) to employ the jobless rate as an additional indicator of the youth labour market situation in their flagship report on youth in the world economy. The OECD also now reports information on this indicator, calling it the NEET (not in employment or education or training) rate. See, for example, O’Higgins (2010b) for a more detailed discussion of why this indicator is useful in the context of youth labour markets. 28/ Indeed, an additional advantage of using the jobless rate here, concerns differences in the nature of Roma (and consequently also non-Roma) samples in different countries. For example, Roma in Slovakia are predominantly found in rural areas, whereas in the Czech Republic, there is a higher proportion of urban residents. Thus, examination of jobless rates removes some of the ‘noise’ from the figures and provides an intuitively more accurate picture of the labour market situation of Roma and non-Roma living in their proximity.

43

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 21: Jobless rates by education in Central and Southeast Europe, 2011 (%)

Source: calculated from the UNDP/WB/EC regional Roma survey 2011.

The correlation between joblessness and education is shown in Figure 21. Data suggests that the ethnicity gap is more pronounced for individuals with higher education. The gap in joblessness between Roma and non-Roma is minimal for persons with no formal education or primary education only but it increases to 16 percentage points for people with post-secondary education. Given a jobless rate for Roma with postsecondary education at this level, it would be hard to persuade a Roma youngster to make the effort and continue with his/hers education. This may also be a part of the reason why highly educated Roma are not commonly perceived as role models. It is also interesting to observe that jobless rates for Roma at each educational level follow almost exactly those of non-Roma with an educational level one level lower. However it is not possible to determine from the available data to what extent this is attributable to prejudice as opposed to the quality of the education attained.

44

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 22: Informal employment rates among Roma and non-Roma in Central and Southeast Europe, 2011 (%)

Source: Calculated on the basis of the UNDP/WB/EC regional Roma survey 2011. The data are shown only for the highest values for each group (Roma or non-Roma).

These data indicate that, even when employed, Roma are disadvantaged in terms of the quality of employment for those who do find work. Informal employment (Figure 22), which results in lower pay and the absence of any kind of employment security, health, and safety/social protection is an indicator of this disadvantage (O’Higgins 2012: 25).29 Data also suggests that in most countries, men have higher informal employment rates than women, although with some exceptions (namely, in the Czech Republic, Hungary, Croatia and Macedonia the opposite is true). The highest gender gap in informality of employment is in Montenegro. 29/ There are difficulties and variations in the definitions of informal employment. In part, this explains the adoption by the ILO of the concept of vulnerable employment which has an unequivocal definition, if not meaning, across countries. In common with the convention in this region, here informal employment is defined as employment for which social contributions are not paid.

45

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

The issue of informality in Roma employment, however, has broader implications and goes beyond its economic dimensions. It is related to the concept of Roma identity as a meta-construct, largely defined through its relationships with others. For centuries, Roma groups have lived side-by-side with the members of majority societies, providing a range of services to the Gadze populations without being integrated in the majorities’ (sedentary) disciplining structures, both clerical and secular. In economic terms, the informality of this exchange was an implicit characteristic, underpinning the pattern of Roma identity (Ivanov 2012: 82-84). With modernization, and the emergence of the modern nation-states, the degree of informality of Roma income generation did not increase in absolute terms – but it did increase in relative terms compared to non-Roma economic agents. These historical roots and their relationship to identity patterns are important to bear in mind today, when addressing the challenges of informality.

Health30 While Roma survey respondents are generally positive in assessing their health status, they also report a higher number of patient and emergency visits, and indicate higher prevalence of disabilities and addictions. The number of in-facility stays (Figure 23) and reported disabilities actually point to higher objective vulnerability of the Roma. On a regional level, the number of insured Roma that reported visiting hospitals during the Figure 23: Inpatient stays, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

30/ This section is based on the findings in Mihailov, 2012. The figures visualise the data derived from the reported experience of the main respondent or a randomly selected member of the household 16+ in a specific health section of the survey.

46

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 24: Access: health insurance coverage, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

past 12 months was 17%, compared to 12% for non-Roma respondents (Mihailov 2012: 21). In nine out of 12 countries, Roma respondents reported an “inability to work due to long-term illness/disability” more than did non-Roma respondents. Exclusion from formal labour market activities also excludes Roma workers (and their families) from regular medical insurance and restricts their access to health services. It also distorts Roma health vulnerability data. Incidents of illness are only captured in statistical data when they require emergency or inpatient interventions. As a result, while the survey data do not show significant differences between Roma and non-Roma in outpatient visits, they show larger disparities in inpatient visits. Once health services are accessed and illnesses are diagnosed, reported health status decreases. This revelation effect is confirmed by a probit analysis showing that the probability that a Roma will report a chronic disease increases with 7% if s/he holds a medical insurance. Age and gender also appear to be leading causes of health status (reported illnesses). While Roma health status seems to have undergone some improvement (both in terms of reported diseases and general self-assessment) in the region since 2004, this may reflect overall improvements in health status, rather than a result of special interventions for the Roma. On the whole, the survey data indicate that the leading causes of Roma health vulnerability are related to a lack of attachment to formal employment and the associated medical insurance programmes (Mihailov 2012: 67). The ethnic gap in insurance coverage is the largest in Moldova, Romania, Bulgaria, Bosnia and Herzegovina and Albania (Figure 24). These countries also have comparatively lower expenditures on health in relations to total governmental expenditures, which proves the positive relationship

47

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 25: Access: affordability of buying medicines when prescribed, 2004-2011 (%)

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011.

between access to health services and the overall living standards in the respective country, including the Governmental contribution to the health expenditures. Financial access to (affordability of ) health services is also significantly lower among the Roma. At the regional level (unweighted averages of pooled Roma and non-Roma samples) 55% of the Roma report instances in the past 12 months when household members could not afford to purchase prescribed medicines compared to 25% of the non-Roma (Mihailov 2012: 37). Physical access to key medical institutions is also lower for the Roma compared non-Roma, although the differences are less explicable than other indicators: 42% of the Roma aged 16+ and 26% of the non-Roma report that during the last year they had experienced a situation when they did not have access to a doctor when needed (Mihailov 2012:39). Most countries have achieved some improvement in that regard between 2004 and 2011 with Montenegro and Hungary registering the biggest improvements (Figure 25). Access to primary medical care and specialised medical testing such as dental tests, blood sugar tests, and radiological services is reported to be particularly limited. While reported access to primary health services coverage (insurance and access to general practitioners) has significantly improved among the non-Roma during 20042011, it declined for Roma in Albania, Romania, and Bulgaria. Particularly striking is the decline of primary health care access in Romania, where 90% of Roma survey respondents reported to having a family doctor in 2004, compared to 49%, stating themselves to be insured in 2011. Overall progress in access to health services among the Roma is lower in countries reporting lower per-capita incomes and lower governmental expenditures on health,

48

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

such as in Albania, Romania, and Bulgaria. The data suggest that the health situation of the Roma in Romania remains at very critical levels. If there is any progress, it would seem to be based on general social advances, rather than on interventions directed at Roma. Whereas only 2% of the Roma give birth at home without professional support, this is higher than the rate reported for non-Roma (Mihailov 2012: 58). Unattended births outside hospitals are higher in Serbia, FYR Macedonia, and Bosnia and Herzegovina. Gynaecological testing is significantly lower among the Roma than the non-Roma, with a less distinctive difference for the younger age groups. The lowest frequency of gynaecological and cervical testing is observed in Albania, Bulgaria, and Romania.

Housing31 The notable gaps in the health conditions of Roma and non-Roma survey respondents are related to differences in housing conditions and access to public utilities and services. Almost one third of Roma households surveyed throughout the region do not have access to improved water sources inside their dwelling (Figure 26). In some counFigure 26: Households without improved water source and sanitation (%), 2011

Source: UNDP/World Bank/EC regional Roma survey 2011.

31/ This section is based on the findings in Perić 2012.

49

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 27: Roma households without access to... (%), 2011

Source: UNDP/World Bank/EC regional Roma survey 2011.

tries this was the case for the majority of Roma respondents (e.g., in Moldova, with 66% of such Roma households; and in Romania, with 72%). Everywhere except in the Czech Republic, the share of Roma households living without piped water inside their dwelling was higher than the respective share of non-Roma. The data for Croatia show the most notable gap between Roma and non-Roma in this respect—35% of Roma respondents (compared to only 4% of non-Roma) were living without these facilities (Perić 2012: 23). Even higher shares of Roma households do not have access to improved sanitation facilities. The highest incidence of this type of deprivation in the region was recorded in Moldova and Romania, at 79% and 78%, respectively. The most notable gap was registered in Bulgaria (62% of Roma do not have a toilet or a bathroom inside their dwelling, compared to 18% of non-Roma living in close proximity). Roma households have more limited access to electricity than their non-Roma neighbours, though the share of those not connected to the grid is lower than of those without access to improved water and sanitation services. The differences in access to these basic infrastructures within individual countries are also revealing. As Figure 27 shows, the biggest problem in all countries is access to sanitation. In some cases (Hungary, or Slovakia) limited access to improved water sources is matched by equal limitations in access to sanitation. In most countries, however, this is not the case. In Bulgaria and FYR Macedonia, for example, the share of those without

50

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 28: Insecure housing of households (%), 2011

Source: UNDP/World Bank/EC regional Roma survey 2011.

indoor toilets or bathrooms is respectively 12 and 11 times higher than the share of those without piped water (Perić 2012: 24). This could reflect the general underdevelopment of communal infrastructure in rural areas (where significant numbers of Roma in those countries live). While lack of access to electricity seems less pronounced than lack of access to water and sanitation, it has equally grave implications for the energy sources used for cooking and heating. Most surveyed Roma households were less likely to use electricity for cooking and heating than non-Roma households living in their proximity. Thus, it is not surprizing to observe that wood and coal were used more frequently in Roma households—despite the health implications associated with the indoor use of solid fuels for these purposes. The survey data also indicate that more Roma households cannot afford to keep their dwellings warm, in comparison with their non-Roma neighbours. Public waste collection services are also less available to Roma, compared to non-Roma in their vicinity. In most surveyed locations waste is not removed for larger numbers of Roma households. Generally, most neighbourhoods of Roma households underwent fewer urban improvement projects, according to survey results (Perić 2012: 29). Habitable shelter is another key concern—considerably larger shares of Roma households surveyed throughout the region live in substandard houses or slums, compared to nonRoma respondents. The share of surveyed Roma population living in insecure housing of this type ranges from 14% in the Czech Republic to 42% in Montenegro (Figure 28).

51

Figure 29: Progress in Roma survey respondents’ access to basic infrastructure, 2004-2011 (in percentage points)

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011.

The physical location of Roma housing is often marginal, in both spatial and social terms. In addition to being contrary to the aims of the Roma Decade, segregated housing also ignores the wishes of the Roma themselves. As the 2011 regional Roma survey data indicate, Roma respondents value living in ethnically mixed housing. However, reallife evidence in much of the region illustrates that many non-Roma do not want to live together with Roma. When asked, roughly three quarters of Roma respondents chose to live in mixed rather than segregated areas. Average preferences per country ranged from 65% of Roma in Moldova to 91% of Roma in FYR Macedonia (Perić 2012: 52). Compared to data from the 2004 UNDP Roma survey, access to improved water and sanitation for Roma households has generally increased throughout the region, but not always, and to different degrees in different countries (Figure 29).32 Whereas progress in access to improved sanitation is evident in all the countries covered by both surveys, progress regarding both improved water sources and improved sanitation has been registered only in Bulgaria, Hungary, and Montenegro (Perić 2012: 65-67).

Integrated proxy for outcome-level poverty The “well-being” dimensions of poverty reflect its multidimensional nature (beyond a lack of monetary incomes). These dimensions have a dual nature—they are both outcomes and determinants of poverty. This is why it is necessary to conceptualize poverty as a complex multidimensional phenomenon. The question is how. It is necessary to track 32/ As the 2004 UNDP Roma survey was not conducted in Slovakia and Moldova, these two countries are not included in the 2004/2011 comparison.

52

POVERTY FROM THE PERSPECTIVE OF FINAL OUTCOMES

Figure 30: Material deprivation index (Roma and non-Roma, 2011)

Source: UNDP/World Bank/EC regional Roma survey 2011.

poverty-related indicators to obtain a comprehensive picture of the challenges. But in order to form an aggregated idea of the progress (or lack thereof), those sector-specific indicators need to be complemented by composite multidimensional indicators. The material deprivation index monitored by EU member states (used also in the Europe 2020 strategy) is one such indicator. Survey data allow the calculation of a modified version of this index for Roma and their non-Roma neighbours, as an index both of “material deprivation” and of “severe material deprivation”.33 Figure 30 suggests that most Roma face “severe material deprivation.” The distance between the two categories of deprivation is substantively larger in the case of non-Roma. 33/ In the EU Material Deprivation index an individual is assumed to experience material deprivation when at least three of the following items are missing in his/her household: (1) the ability to pay rent, mortgage, or utility bills, (2) the ability to keep the house adequately warm, (3) the ability to face unexpected expenses, (4) the ability to regularly eat meat or proteins, (5) to go on holiday, (6) a TV set, (7) a washing machine, (8) a car, (9) a telephone (mobile or fixed line). When at least four of these items are lacking in the household, its members are said to experience severe material deprivation. In the index used in this paper the following six items were considered: having a TV set, a washing machine, a car, a telephone (mobile or fixed line), one bed per person and a computer. A person was considered “materially deprived” if the household was missing 4 out of the 6 categories.

53

4

LOOKING BEYOND MATERIAL WELLBEING

While the material deprivation index offers useful information, it reflects only the material aspects of well-being.

From “basic needs” to “human development” The idea behind the human development concept is simple, intuitive and at the same time extremely rich. From the human development perspective money is merely the means to expand “people’s freedoms to live long, healthy and creative lives; to advance other goals they have reason to value” (UNDP 2010: 22). The concept of “human development” is not the first attempt to look at poverty through a human perspective. In 1976, the ILO published its report “Employment, growth and basic needs” which launched the idea of “basic needs” at the level of international development policies (ILO, 197634). These needs concern the material standards of decent living, access to education, health care and housing availability. These are important areas of human development as well, and the two concepts are often confused. The difference between the two is in the presence (or absence) of agency, the freedom, the desire and the ability of the individual to take responsibility for his/her own destiny. In 1976, the ILO did not introduce an integrated indicator for measuring progress in the area of “basic needs”—UNDP did this in 1990 with the human development index. The HDI has since emerged as an alternative to GDP (and hence, monetary metrics) as a measure of human progress. However, even the HDI does not explicitly address the agency. So—a true paradox—the UNDP index of “human development” is actually an index of “basic needs.” Another problem is related to the question what is measured. The HDI is not a measure of deprivation. It is an indicator of the achievement of progress (and not of poverty or other development challenges). The differences between these two aspects are not just terminology. To a certain extent one might “reverse” the HDI and use the distance of the achieved level from the maximum values of its components as a measure of 34/ Of course, the concept of “basic needs,” as formulated by the ILO, not only goes beyond money as a measure of well-being and its components are the only ones. Detailed analysis of the approaches in this going on-line, however, is beyond the scope of this analysis.

54

development shortfall. However, one cannot reverse the determinants of success to explain failure. In response, UNDP in 1996 introduced a “human poverty index” to measure deprivations along a number of dimensions. HPI-1 was comprised of three indicators: survival (percentage of individuals with a life expectancy lower than 40); deprivation of knowledge (expressed as a percentage of illiterate adults), and deprivation of decent living standards (a simple average of three basic variables: the share of the population without access to drinking water; the share of the population without access to health services; and percentage of underweight children aged less than five). As the HPI-1 was intended to reflect the challenges in developing countries, the HPI-2 was introduced in 2007 to address development challenges in wealthier countries. The survival dimension measured the percentage of individuals with a life expectancy lower than 60; deprivation of knowledge was defined as the share of functionally illiterate adults; while deprivation of living standards was defined via the monetary poverty rate (with 50% of the average national per-capita income applied as a threshold [UNDP 1997]). HPI-2 also considered a fourth dimension (“social exclusion”) using the long-term unemployment rate as a proxy.

From “basic needs” to social exclusion Social exclusion is however broader than unemployment. UNDP’s 2011 regional human development report (UNDP, 2011) argues that income-based poverty measures are not able to capture the depth and breadth of the deprivations in the region today. The report defines social exclusion as both a process and an outcome. As a process it pushes certain individuals to the social margins and prevents them from realizing their full potential; as an outcome, it denotes the status and characteristics of excluded individuals. Outcomes (results of past deprivations) function as determinants in the current moment and in the future. The key contributions of the report were approaching social exclusion from the perspective of individual characteristics, and the introduction of the social exclusion chain. The report departs from traditional group-based approaches and assumes that each individual has a number of individual characteristics that can put him or her at risk of social exclusion. These are the social exclusion risks that can be related to gender, ethnicity, language, religion, age, sexual orientation, religious beliefs and disability, as well as those linked to status (income, health, employment, education, resources, opportunities, and assets). But not all individual risks result in social exclusion. Whether social exclusion occurs depends on the interactions of risks with a set of ‘drivers’ that can be structural, behavioural, or policy-related. Social exclusion is seen as the outcome of multiple and mutually reinforcing deprivations in one or more dimensions (UNDP 2011: 11-13).

55

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

The report introduced a social exclusion index35 based upon the Alkire and Foster methodology of multidimensional poverty monitoring (Alkire and Foster 2007). The data used for computing the index were derived from a regional social exclusion survey (based on a set of nationally representative household surveys) implemented in six countries.36 In one of them (namely, Serbia) it was conducted on separate samples of Roma and internally displaced persons (IDPs). This survey finds that 86% of Roma and 56% of IDPs are socially excluded, compared with 19% for the general population. The intensity of exclusion is also found to be much greater. The average number of deprivations experienced is 15.2 for socially excluded Roma and 13.2 for socially excluded IDPs, compared with 10.8 for the general population sample. Furthermore, for any given cutoff (threshold), significantly greater numbers of Roma and IDPs are socially excluded as compared with the general population (Figure 31). This is true even when income indicators are removed: Roma and IDPs are also more deprived in the non-material dimensions of social exclusion. These data indicate that factors like unemployment and disability significantly increase the risk of social exclusion among Roma and IDPs. For example, the share of socially excluded is 92% among unemployed Roma and 97% among those with disabilities (compared to 86% for Roma on average). The share of socially excluded is 76% among unemployed IDPs and 73% among IDPs with disabilities. While it is not surprising that social exclusion increases for Roma and IDPs who face several overlapping risks, disability has a lower impact on the exclusion of IDPs. This suggests that disabled IDPs might have slightly better access to social safety nets than disabled Roma. Analysing the deprivations experienced by Roma and IDPs as compared with the general population yields further insights. For example, in the “access to social services” dimension of exclusion, deprivations contribute in varying magnitudes across the three groups. While deprivations in education and health contribute more significantly to exclusion for Roma, they are less pronounced for IDPs and the general population. The picture is similar when looking at the three groups in the other two dimensions of exclusion. 35/ The social exclusion status is assessed based on deprivations reported along three dimensions and captured by 24 indicators (eight indicators per dimension). In the first dimension (economic exclusion) indicators reflect deprivation in income and basic needs; employment, financial services and material assets; amenities that households need but cannot afford; and dwelling size. Indicators in the second dimension (exclusion from social services) encompass education and health services, as well as public utility services. The third dimension (exclusion from civic and social life) covers deprivation in terms of political, cultural, and social networks, as well as reflecting diminished opportunities for social and civic participation. 36/ The survey was conducted in 2009 among 2,700 persons in each of six countries: Kazakhstan, Moldova, Serbia, FYR Macedonia, Tajikistan, and Ukraine. In Serbia the survey sample included internally displaced persons as well as members of the Roma minority. It employed the same methodology for all countries, thus permitting cross-country comparisons. Qualitative information from focus-group discussions and individual interviews with vulnerable groups complemented the quantitative data and provided valuable insights into the experiences of socially excluded persons that are difficult to capture through traditional survey techniques

56

LOOKING BEYOND MATERIAL WELLBEING

Figure 31: Vulnerable groups and social exclusion

Source: Regional Social Inclusion Survey (UNDP 2011).

From “Human Development” to “Multidimensional Poverty” UNDP dropped the HPI in 2010, and adopted the multidimensional poverty index as a measure of deprivation complementing its human development index as a measure of progress achieved. The MPI denotes overlapping deprivations at the household level across the same three dimensions as the HDI (living standards, health, and education). It shows the average numbers of poor people and the deprivations with which poor households contend.37

37/ http://www.ophi.org.uk/multidimensional-poverty-index/mpi-faqs/

57

5

ROMA POVERTY FROM A MULTIDIMENSIONAL PERSPECTIVE

The multidimensional poverty approach is particularly appropriate to addressing the issue of Roma poverty, precisely because of the multidimensional nature of the deprivation and marginalization that Roma are faced with. Roma poverty is not just a lack of financial resources, unemployment, sub-standard housing, or poor access to social services. It is a combination of all these factors, which are both the reflection of prior experiences of exclusion and the determinants of future deprivations—reinforcing the vicious cycle of poverty. This vicious cycle is reinforced by prejudice and discrimination, specific behavioural traits, limited opportunities to participate in political processes, etc. The multidimensional nature of Roma poverty calls for a “human development” and not just a “basic needs” approach. It should also integrate the reduction of material deprivation with increasing agency and the achievement of fundamental civil rights. Unemployment, social exclusion, and marginalization are linked with (and are mutually reinforcing) discrimination, anti-Gypsyism, limited access to justice, and segregation. In order to capture these aspects of Roma deprivation, a multidimensional poverty index has been developed and tested following the standard Alkire and Foster (2007) methodology. This index integrates important aspects of human poverty and reflects the specifics of Roma exclusion appropriately.

The construction of the index This index reflects the status of the individuals (and their characteristics) living in households (with their characteristics) and facing a number of deprivations. It combines 12 equally weighted indicators which reflect their status in six critical dimensions based on a human development perspective (basic rights, health, education, housing, standard of living, and employment). Table 2 summarises the specific indicators, dimensions, and areas as well as the information required for the individual indicators (individual or household). The status of the individual in each dimension is tracked with two indicators per dimension. The first three dimensions cover “human capabilities” of which basic rights, education, and health emerge as particularly important. Obviously, the entire palette of fundamental rights is far richer than the two indicators. It includes the right to work, protection of individual security, etc. However, both indicators reflect the presence or absence of the necessary conditions for the realization of other fundamental rights. The

58

Level of observation

Having an ID – yes/no (personal document, birth certificate etc.)

I

Discrimination

HH member lives in a HH where a member has been discriminated against while looking for a job

P

Health (1/6)

Disability status

A household member having a disability – yes/no

I

Limited access to medical services

Any HH member living in a HH responding “yes” to the question “were there any periods in the past 12 months when you couldn’t visit a doctor when you needed?”

P

Highest completed education

For adults: any HH member above schooling age who hasn’t completed primary education or lower secondary

Basic rights (1/6) Employment (1/6)

Standard of living (1/6)

Housing (1/6)

Human capabilities Material wellbeing

Indicators Civil status

Dimension and weight

Criterion of deprivation and threshold

Education (1/6)

Area

Table 2: Dimensions and indicators of the “Roma multidimensional poverty index”

I

For children: children in school age who are not in school

Self-declared illiteracy rate

Any HH member stated as unable to read and write

Access to basic infrastructure

A composite indicator –any HH member living in a HH without two of the three (toilet or bathroom inside the house; running water; electricity)

Shares of the population not Any HH member living in “ruined houses” or “slums” having access to secure housing

I

H

Any HH member living in a HH that experienced that Extreme poverty in the past month somebody ever went to bed hungry because they could not afford enough food for them

H

Access to various HH amenities

Any HH member living in a HH, which doesn’t possess four of six categories falling in the “Material deprivation” index

H

Unemployment

Any HH member living in a household with none of the adult HH members employed (16+)

H

Lack of working experience

Any HH member living in a HH in which the HH head or his/her spouse has no working experience

H

Level of observation of the respective indicators: I – individual status of each household member P – the experience and perception of the main respondent extrapolated to all household members H – the status (vulnerability) of the household along a certain parameter extrapolated to all household members

59

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 32: Multidimensional poverty rate and its composition for Roma (nonsevere and severe, bars left scale) and the value of MPI (right scale) (%)

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011.

understanding of “capabilities” is also slightly different from the traditional definition adopted in the capability approach reducing them to personal characteristics and excluding material aspects of capability. But since the individual dimensions are equally weighted, this distinction has no material significance for the index. The second group covers the major aspects of “material well-being.” The index is calculated on the basis of the “individual status of each member of the household”. This status reflects either the personal characteristics of the individual in question, or the condition of the entire household shared by all its members and extrapolated as an individual parameter to each household member. In determining multidimensional poverty status, one cut-off line with two levels was applied: one for “multidimensional poverty” and one for “severe multidimensional poverty”. People experiencing 5-7 deprivations were considered “multidimensionally poor”; those experiencing eight or more deprivations were considered “severely multidimensionally poor”. Unlike the Alkire-Foster MPI methodology, no cut-off within dimensions was applied, because of the limited number of deprivations in each dimension (2) and the dichotomous nature of most variables.

60

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

Figure 33: Multidimensional poverty rates (total) for Roma and non-Roma, 20042011 (%)

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011.

This methodology allows for integrating in a single index the poverty rate (the share of people experiencing five or more deprivations) and the severity of poverty (the average number of deprivations experienced by those in poverty). The MPI is the share of the multidimensionally poor multiplied by the average number of deprivations they experience. Figure 32 presents the results for the countries covered by the survey. As Figure 32 shows, the multidimensional poverty rate of Roma has decreased substantially between 2004 and 2011 in most countries (it increased in Albania and Serbia). This is clearly an achievement. In most of those countries witnessing improvements, however, they resulted primarily from declines in the number of Roma in the “poor” category. “Severe poverty” decreased significantly only in Bosnia and Herzegovina and Romania; in Albania, it even increased. Have declines in multidimensional poverty headcounts been shared equally? The data shown in Figure 33 suggest that the answer in most cases is “no.” While in most countries multidimensional poverty among Roma has declined (with the exception of Albania and Serbia), the opposite is true for non-Roma. It declined only in Romania—and with a magnitude similar to the decline among the Roma. In Albania it increased even with a higher magnitude. In the rest of the countries the increase of non-Roma multidimensional poverty was modest but still observable.

61

The data in Table 3 suggest that the average number of deprivations experienced by Roma and non-Roma households falling into the category “poor” declined during 2004-2011 in most countries. The change was not uniform across countries and groups, however. It increased for both groups in Albania and Montenegro and for non-Roma in Macedonia. In some countries the two trends were diverging – Serbia for example experienced an increase in multidimensional poverty rates and decline in average number of deprivations for both Roma and non-Roma suggesting that the overall share of people living in poverty increased but they have become “less poor” overall. Montenegro experiences the opposite in the case of Roma – the decline in overall poverty rates was matched by an increase (albeit slight) of the average number of deprivations. Romania marked the most significant decline in average number of deprivations among Roma and this reflected in the decline in overall multidimensional poverty rates. The diverging trends in regards poverty rates and severity of poverty suggest that a more nuanced analysis of the phenomenon is needed for effective poverty reduction

Table 3: Average number of deprivations 2004

2011

Difference

Roma

6.22

6.32

0.11

Non-Roma

5.25

5.69

0.44

Roma

6.51

6.13

-0.37

Non-Roma

6.02

5.52

-0.50

Roma

6.17

6.05

-0.12

Non-Roma

5.80

5.50

-0.30

Roma

6.23

5.90

-0.33

Non-Roma

5.50

5.50

0.00

Roma

6.02

6.08

0.06

Non-Roma

5.20

5.94

0.74

Roma

6.08

5.99

-0.09

Non-Roma

5.53

5.74

0.22

Roma

6.77

6.18

-0.59

Non-Roma

6.06

5.72

-0.33

Roma

6.30

6.21

-0.09

Non-Roma

6.30

5.65

-0.65

AL

BA

BG

HR

ME

MK

RO

RS

62

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

policies. In some cases the decline in recorded (and reported) poverty rates may be misleading unless the bigger picture is taken into consideration. An important part of this bigger picture is local specifics and the type of deprivations individual household face. Factoring those specifics in is a challenge for effective multidimensional poverty reduction benefitting Roma and non-Roma alike.

Drivers of multidimensional poverty The data in Figure 34 show which deprivations contribute most significantly to these poverty trends, and suggest which areas might be prioritized in poverty reduction/ Roma inclusion efforts. In most countries the average number of deprivations experienced by Roma decreased slightly during 2004-2011. Those who are multidimensionally poor have remained equally poor—but in different ways. The contribution of shortcoming in education and living conditions to multidimensional poverty has declined, while that of shortcomings in fundamental rights and labour activity has increased. The number of deprivations non-Roma face is on average nearer to the cutoff of nonpoor. The structure of deprivations is similar to that of the Roma with one difference: the non-Roma poor have less of a problem with the education or housing dimension but more in the health dimension. Figure 34: Contribution of individual dimensions to multidimensional poverty of Roma (2004-2011)

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011.

63

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 35: Imagine… How much multidimensional poverty would go down if the respected deprivations are overcome in full Individual deprivations’ contribution to multidimensional poverty reflected in the value of the Multidimensional poverty index for Roma and Non-Roma 1

2

3

4

5

6

7

8

9

10

11

12

64

Source: UNDP/World Bank/EC regional Roma survey 2011.

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

Tracking the impact of individual sector-specific policies on poverty is a major contribution of the proposed methodology. It makes possible linking the individual sectorspecific interventions to the overall poverty reduction outcome. In that way policy interventions can be prioritized and allocation of resources devoted to poverty reduction can be optimized. The structure of the average number of deprivations and related changes over time provide a very crude picture of how different factors affect poverty. The fact that a certain dimension constitutes 30% of the overall average number of deprivations does not mean that reducing to zero the numbers of people deprived in that dimension would reduce the multidimensional poverty rate by 30%. This is why for estimating the possible effect of poverty alleviation policies, instead of the poverty headcount, it is more appropriate to use the multidimensional poverty index, which is the share of the multidimensionally poor multiplied by the average number of deprivations they experience. Figure 35 visualises the results of this estimation. It shows the contribution of individual deprivations to the overall MPI value – and, reversing the logic, what would happen to multidimensional poverty should the specific deprivation be eliminated. In other words, it shows what the impact on multidimensional poverty of a specific policy (and the investment necessary) might be. This makes both the figure and the MPI truly policy oriented. The “weights” for the Roma are much heavier than for the non-Roma because the Figure depicts the possible implications of successful policies on the multidimensional poverty index. The multidimensionally poor non-Roma face a similar number of deprivations as the Roma do but their share in the non-Roma population is much lower. Hence the MPI value for non-Roma is also much lower. Among the Roma the largest factors that appear in contributing to this weighted index are the employment indicator, the absence of extreme poverty (i.e., not having gone hungry at least once in the past month), the education indicator, and the material deprivation indicator. Across countries these appear to influence a large portion of the MPI Roma poor. This makes sense as these are usually the factors most often referred to by policy makers as potentially reducing poverty, especially employment and education. Some indicators such as access to basic amenities (piped water, electricity, and sewage) have a differentiated impact; they appear to have a greater impact on MPI poverty in Moldova and Romania than in Montenegro and Bulgaria. A few indicators (e.g., the absence of informal employment—which is not very prevalent among the poor, and does not lead to many people exiting MPI poverty) appear to have hardly any effect on the weighted MPI. These results should be taken with caution as a given policy instrument may affect more than one deprivation, even causing one to increase while another could decrease. A causal analysis of the different components’ impact on the weighted MPI is beyond the scope of this paper, but should be considered for future research.

65

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Table 4: Individual characteristics’ contribution to the risk of poverty Estimates

Roma

Female

Age

Age squared

Household Size

Pre-school

Bad Health

Rural

Roma Dominant

Full Sample

Roma

Non-Roma

MPI poor+MPI severe poor

MPI poor+MPI severe poor

MPI poor+MPI severe poor

0.00342

0.00768

-0.00597

0.00412

0.00505

0.00409

-0.00681***

-0.00665***

-0.00381***

0.00055

0.000724

0.000457

7.51e-05***

7.34e-05***

3.90e-05***

0.00000716

0.00000978

0.00000527

0.0137***

0.0151***

0.00706**

0.00237

0.0028

0.00301

-0.129***

-0.156***

-0.0379***

0.00922

0.012

0.00743

0.171***

0.153***

0.168***

0.0104

0.012

0.0157

0.0342***

0.0411***

0.00898

0.0117

0.0143

0.00866

0.0368***

0.0409***

0.0149

0.0111

0.0137

0.0109

0.303*** 0.00919

Notes: The dependent variable is equal to 1 if the individual is either MPI poor or MPI severely poor. The results are the marginal effects using STATA’s mfx2 command. All probits are clustered at the primary sampling unit. Standard errors are given below the marginal probability coefficients. Difference significant at the 10% level *, 5% level **, and 1% level ***

The results of the probit analysis shown in Table 4 offer another way to measure the effects of certain variables on MPI poverty. In this case exogenous variables are tested against the headcount of the MPI poor and severe poor (MPI equals 1 if an individual is either MPI poor or MPI severe poor, and 0 otherwise). Marginal effects for each variable are calculated and if they are positive this means that as they increase, they increase the probability of being MPI poor.

66

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

The results suggest that being Roma increases the probability of being MPI poor by 30.3%, but that being female has no significant effect. Age affects the risk of poverty differently along the life span. As people get older they are less likely to be poor (the probability declines by 0.7 percentage points per year)—most probably because adulthood is associated with a reduced risk of youth unemployment. However, that probability begins to go up after the person reaches 45, suggesting a link to old-age poverty (associated with declining chances of productive employment). Family size matters: each additional person in the household increases the chance of being MPI poor by 1.4 percentage points such that a household with five people has a 5.6% higher chance of being poor than a single-person household. Attending preschool matters dramatically—it reduces the likelihood of being MPI poor by 12.9%. Even more powerful is the impact of health—having bad or very bad health increases the chance of being poor by 17.1%. Rural/urban division also matters—being from a rural area increases the chance of being poor by 3.4%. Interestingly enough, living in a Roma-dominant area increases the chance of being poor by only 3.7%. This may mean two things. One explanation might be that even when segregated, Roma households still do not constitute a “separate universe” and closely interact with non-Roma communities. A second explanation is that, even when living in a mixed area, Roma are still treated with prejudice and cannot benefit fully from the opportunities of integration. Given the huge impact of “being Roma” on the risk of falling into poverty (more than 30%), the second interpretation seems more likely to be the case.

Correlates with other dimensions of poverty Analysing the status of the individuals in areas that are crucial from a human development perspective in the context of multidimensional poverty status offers additional insights into the relationships between the various aspects of poverty. For that purpose socioeconomic and human development indicators were calculated for subgroups defined by poverty status. This approach allows us to draw more robust conclusions at the regional level. Given the diversity among Roma within and across countries, a regional “Roma” sample may be subject to reasonable criticism. However a sub-sample based on multidimensional poverty status—or more precisely, disaggregated by the similarity of poverty characteristics—allows the grouping together and analysis of households that find themselves in similar life situations and facing similar challenges. Thus the conclusions in regards to their status and poverty drivers may be applicable beyond the specific country context and complement country-specific analysis. Figures 36-42 show some examples of what such an analysis might yield.38 38/ Because it is a composite measure, it’s logical to expect that multidimensional poverty will be correlated with some of the indicators. This is why only those indicators that are not included in the structure of the MPI are considered here.

67

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 36: Correlation between multidimensional poverty and achieved level of education of adults (2011) (%) Share of household members ending educational years with the respective ISCED level of education, (age 25-64)

Source: UNDP/World Bank/EC regional Roma survey 2011.

Education A higher achieved educational level goes hand-in-hand with lower multidimensional poverty (Figure 36). Education in that regard is both an outcome and a driver of deprivation. The similarity between the “severely poor” Roma and non-Roma is worth noting. It suggests that the deeper the poverty, the less significant are its ethnic determinants. But this does not mean that the decline in poverty translates symmetrically into higher education. The negligible share of non-poor Roma with ISCED3 and 4+ indicates that ethnic factors have a stronger impact on educational achievement for non-poor groups. The correlation between multidimensional poverty and education is much stronger for the younger generation (aged 20-24, i.e., those who have completed their educational years) that for other age cohorts. As Figure 37 shows, severe multidimensional poverty is clearly correlated with low educational achievement of the young people living in deprived households. Also, the similarity between Roma and non-Roma in similar poverty status no longer holds true. Most probably this is due to the decreasing influence of the former centrally-planned socialist system with its obligatory (and enforceable) primary and secondary education enrolment policies. As this is no longer the case, group-related specifics and horizontal inequalities are much more pronounced. Figure 38 illustrates even more explicitly the relationship between poverty and “poverty in the making.” Low school enrolment rates lead to lowered prospects in life later

68

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

Figure 37: Correlation between multidimensional poverty and achieved level of education of youth (2011) (%) Share of household members ending educational years with the respective ISCED level of education, (age 20-24).39

Source: UNDP/World Bank/EC regional Roma survey 2011.

Figure 38: Gross enrolment rates in respective education levels, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

39/ There are not enough MPI severe poor among the non-Roma so the category is left out, otherwise results would be misleading.

69

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 39: Average years spent in education by poverty status and age group, 2011

Source: UNDP/World Bank/EC regional Roma survey 2011.

on. The data point to the trans-generational nature of poverty-induced vulnerability. People living in severe poverty have the lowest enrolment rates at all levels of education, with all the associated implications for their future employability. Interestingly, the distance between “poor” and “non-poor” is lower than between “poor” and “severely poor”. Ethnically-related differences are observable at both school levels, but not for pre-school. This might be due to the differences between countries that are lower in the case of compulsory and upper-secondary schooling. Finally, the differences between Roma and non-Roma are smallest at the level of compulsory schooling (7-15 years) and increase sharply for the upper secondary educational level. Average years spent in education is another interesting indicator that, even with caveats, is close to an “educational outcome”.40 Figure 39 shows the differences both between poverty and ethnically-defined groups. Roma lag behind on both counts. The multidimensionally poor non-Roma spend on average 1.67 the time at school than do the multidimensionally poor Roma. The respective ratio for the non-poor is 1.42. It should be noted that the groups considered have similar average numbers of deprivations. Generationally, the younger Roma seems to be spending slightly longer at school than their parents. This is not the case for the multidimensionally poor non-Roma, however. The differences between countries are shown in Figure A in the Annex.

40/ The ultimate purpose of education is gaining meaningful knowledge. “Average years spent in education” (similarly to “enrolment rates”) are a means to this end. They increase the probability of gaining knowledge, but definitely do not determine knowledge outcomes.

70

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

Employment Predictably, multidimensional poverty goes hand in hand with employment vulnerability. Figure 40 shows unemployment rates by multidimensional poverty status. The relationship is clear enough to not require any in-depth analysis. Table 5 on the other hand shows the average duration of unemployment spells of those who are currently out of work. The higher vulnerability of the multidimensionally poor in this regard is not

Figure 40: Unemployment rates by multidimensional poverty status and age group, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

Table 5: Average length of unemployment for those that have worked before and are currently unemployed (as per the ILO definition) in the age group 15-64 Group by ethnicity

Roma

Non-Roma

Group by poverty status

MPI Severe Poor

MPI Poor

MPI NonPoor

MPI Severe Poor

MPI Poor

MPI NonPoor

Years

8.07

6.52

5.00

6.29

5.59

4.83

As % of the non-Roma unemployment spell

128%

117%

104%

100%

Source: UNDP/World Bank/EC regional Roma survey 2011.

71

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 41: Share of those who have never worked in the unemployed population in the respective age groups, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

Figure 42: Different aspects of informality, 2011 (%) The shares of respondents aged 16-64 who would rather have secure employment having to be at work 8 hours a day 5 days a week to those preferring irregular employment leaving them free to manage their own time.

Source: UNDP/World Bank/EC regional Roma survey 2011.

72

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

surprising. Worth noting is the similar duration of unemployment for multidimensionally non-poor Roma and non-Roma (0.27 years). The distance in that regard increases with the deterioration in multidimensional poverty status, reaching 1.78 years for the severely poor (or 128% of the severely poor non-Roma unemployment spell). The differences between countries are shown in Figure B in the Annex. Figure 41 depicts the real seriousness of the unemployment problem—youth unemployment. There are virtually no differences between the shares of unemployed with no working experience between different poverty groups, regardless of ethnicity. Informal employment is rampant among the multidimensionally poor, with only minor differences between the two poverty groups. 97% of the employed Roma and 100% of the employed non-Roma from the “severely poor” group work without contracts. These shares among the “multidimensionally poor” are 88% and 81%, respectively. The share of informally is substantially lower among the non-poor, 41% among the Roma and 21% among the non-Roma, respectively (Figure 42). When choosing between low-paid, secure employment versus having better-paying but insecure, irregular employment, all groups regardless of their ethnicity or poverty status prefer the first option (lowpaying security over higher-paying risk). Between 80% and 87% of the respondents aged 16-64 would make this choice. This share is slightly lower among the severely poor non-Roma (71%). Views on informality show significant cross-country variation, however (see Figure C in the Annex).

Behavioural patterns and aspirations The multidimensional poverty profiles support an in-depth analysis of different socioeconomic and behavioural characteristics and their correlation with the level of poverty. The data in Figure 43 (p. 74) illustrate the extent to which certain behaviours are acceptable (the share of respondents in the respective poverty category who responded “yes, entirely permissible”). These data indicate that extreme poverty—regardless of ethnic affiliation—is correlated with lower thresholds for intolerance towards petty household theft for the sake of one’s own survival, or towards tax evasion (the results for tax evasion tolerance match those for stealing food). Although “correlation” does not imply “causality”, these data strongly suggest that reduction in extreme poverty would result in the reduction of such incidents, about which many non-Roma communities are highly sensitive. The differences between ethnic groups that emerge for the other two groups (non-severely poor and non-poor) may perhaps be explained by the fact that the issue of hunger remains more hypothetical for non-Roma than for Roma. Also, significant differences between countries should be noted (Figure D in the Annex). Again, these data should be properly contextualized. The pattern of cohabitation between Roma and the Gadze and the experience of discrimination has shaped Roma survival strategies. Gadze were seen as fundamental source of income – but also a barrier to free access to needed goods, which led Roma to resort to shmekeria the pe-

73

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 43: Share of respondents who said that stealing food is acceptable if it prevents a family from going hungry, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011. Based on the question “Please tell me is stealing food if your family goes hungry acceptable?” (Those choosing “fully acceptable” rather than “somewhat acceptable” or “not acceptable)

Figure 44: Types of employment aspirations, 2011 (%) Shares of respondents who prefer living on social assistance with problems making ends meet but with no particular effort, by respective age groups

Source: UNDP/World Bank/EC regional Roma survey 2011. Based on the choices by the respondents between the two options (“Live on social assistance with problems making both ends meet but with no particular effort” and “Have higher standards of living but working hard to earn my living“) on the question “People often have to choose between different options in life. Which one would you choose if you face each of these options?”

74

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

culiar cunningness, “submissive demeanour where cajolery and deception became a technique of survival” (Biro et.al. 2013: 21). Shmekeria is not just a pattern of income generation; it is also tool for obtaining a moral advantage or compensation for the subordinate position in the Gadzo society. It inevitably leads to a degree of relativity in societal norms and the variance in the acceptability/unacceptability of certain patterns of interaction in regards one’s own group and that of the opponent.41 And this relativity is not unique for Roma/non-Roma – similar patterns are common in the interactions between other discriminated/discriminating populations defined by ethnic or religious criteria and living in proximity (like in the case of Jews/Christians or Arabs/Israelis). This relativity however is asymmetrical. Petty crimes are indeed linked to shmekeria but are not an exclusively Roma speciality. The shmekeria of Roma is simply more visible and more annoying for the non-Roma. This over-exposure leads to a paradoxical indifference to large-scale white-collar crime like shadow privatisation, tax evasion, financial fraud etc. – all with economic theft of a magnitude incomparable to that of the Roma shmekeria. The majorities however remain amazingly complacent of such cases (UNDP 2002: 71-72). The data also illustrate differences in employment aspirations by poverty group (Figure 44). The multidimensionally poor appear more demotivated than the non-poor. However, there are no significant differences in the aspirations of youth (ages 15-24) and their parents (ages 25-65). Finally, the data reveals the important impact of religious institutions like churches and mosques on norms, behavioural patterns and ultimately – the poverty outcomes. These effects come through their influence on economic, social and psychological parameters. For example, the data shows a robust positive and significant effect of churchgoing on education – parents’ churchgoing increases the educational completion rate of their children by a little more than 50% compared to Roma parents who do not attend church or mosque at least once a month. Although the specific channel for this result is not definitively known it is undoubtedly related to the church’s influence on social norms, networks and specifically on parent’s desires to further educate their children. Interestingly enough, the impact occurs among the Roma and not among the Non-Roma. This result lends credence that it is the church’s effect on the specific norms of the Roma (in this case the demand for education of their children) rather than some alternative effect on Roma and non-Roma alike (Kagin 2013). These findings also suggest that governments may not be the most appropriate tool to address behavioral issues. Probably most excluded individuals do not appreciate the conditional aspects of some educational grants or other ways to be forced into participation in education like through taxation schemes. Church and mosque participation is voluntary in most

41/ As the prominent Roma activist Nikolae Gheorge writes, “some Roma lead their horses to the fields of the Gadje, steal their potatoes, while children make off with the peasants’ hens – this is part of our internal knowledge... This is not so far removed from the kind of liberal, individualistic attitude that puts winning before behaving ethically and results in switching political alliances or abandoning values to further one’s own or family business” (Gheorghe 2013: 98).

75

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 45: Educational aspirations of multidimensionally poor parents (Roma and non-Roma) for their children (boys and girls), 2011

Source: UNDP/World Bank/EC regional Roma survey 2011. Based on the question “What do you believe is a sufficient level of education for a child (boy and girl)?”

cases and since the institutions focus on behavioral aspects they may be the most appropriate to address norms. It is also worth revisiting the linkages between multidimensional poverty and aspirations. Parents’ desires for their children’s life opportunities can be a good proxy for parental aspirations for their children. Figures 45 and 46 show the levels of education that the adult respondents deem sufficient for their children, by poverty status and sex. It is remarkable that there are no major gender differences within ethnic groups— respondents in both poverty groups show similar preferences for boys and girls. This may be evidence of an achievement of some gender equality policy goals, but further research is needed. Huge differences however can be observed between Roma and non-Roma. While the majority of non-Roma aspire to ISCED4 or higher level of education for their children (57% and 56% for girls and boys respectively), the similar majority of Roma (54% and 57%) deem ISCED3 as sufficient educational level for their children. This seems to be a typical example of “valuation neglect”, when people’s desires are constrained by what seems possible (Luggieri Laderchi et al., 2003: 14).

76

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

Figure 46: Educational aspirations of multidimensionally non-poor parents (Roma and non-Roma) for their children (boys and girls), 2011

Source: UNDP/World Bank/EC regional Roma survey 2011. Based on the question “What level of education is sufficient for your child?”

The missing dimension of agency As noted at the beginning of this chapter, a major dimension of human development that of agency - is missing, both in the standard human development indicators (HDI in all of its varieties) and in the multidimensional measure proposed in this paper, as well. This is due to conceptual reasons and data deficits. Applying the concept of agency is a major challenge and generating robust data to populating appropriate indicators of agency is an even bigger one. The standard approach uses sample surveys to capture the degree to which respondents feel that they have “control over their lives,” or their perception of the optimal balance between the role of the state and that of the individual in achieving personal success and realizing one’s aspirations (see Alkire 2005; Ibrahim & Alkire 2007; Samman & Santos 2009). Indeed, these aspects are critical to the perception of agency, but such data exist only for national level samples and rarely for group-targeted research. What is more important, such questions are not particularly useful if detached from the ef-

77

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

fects of complex internal group dynamics, local, social and political context, as well as the patterns of interaction among various stakeholders. Historical experience of prejudice and discrimination also powerfully affects the range—and limits—of individual aspirations. All of the above contributes to a “status bias” in human development indicators at the expense of agency: the tendency to quantify and monitor the status and the magnitude of deprivations in various dimensions rather than the opportunities people have (or lack) to reach their desired status and realize their aspirations. Achieved status may be seen as a proxy of opportunities, but only to a limited extent - and this is definitely not the case in marginalized communities experiencing discrimination. For this reason the issue of agency is critical to analysing Roma poverty and the possible approaches to addressing it. Agency may be defined as the aspirations of an individual (or a group), matched by the resources and opportunities required to reach those aspirations. Seen from this perspective, it would be hard to find a group more in need of explicit agency focus than the Roma, who face a vicious circle of high levels of deprivation in virtually all spheres of life that are mutually determining and reinforce each other (UNDP 2002: 42). These deprivations lead to low aspirations that can be met through “low agency strategies” and thus additionally fuel the cycle of exclusion, replicating its patterns over generations. Roma life takes place in an “agency-hostile” context, with powerful interests vested in keeping the Roma in a subordinate status and preventing them from taking their destinies into their own hands without the need of permanent support from various intermediaries from within and outside their communities.42 Roma and pro-Roma civil society organizations (which are different from “civil society”) experienced dramatic growth over the last two decades and were instrumental in articulating human rights violations, social exclusion, or territorial segregation of Roma. However, they still operate primarily at the intermediary level, between international donors and the people on the ground. The vast majority of marginalised Roma communities remained untouched by and detached from the operation of CSOs that are often co-opted by mainstream structures. By not being present in Roma communities, these “intermediating NGOs” in fact disempower local Roma communities, limiting their opportunities to initiate change (Kóczé 2013: 56). Of course, one immediate determinant of the lack of agency is poverty (Sen 1985). The difficulty is that addressing multiple deprivations and nurturing agency is not a se42/ Nikolae Gheorghe puts it explicitly in one of his latest essays before passing away: “What has made informal or ‘traditional’ Roma leaders like the vajda or bulibaşă so powerful recently? What has helped them re-emerge as potential partners for policy-makers and national, regional and local authorities? Why have they gained legitimacy, especially when European agencies and public opinion demand ‘good practices’ and ‘concrete measures’ with the expectation of quick results, which in reality may be no more than window-dressing? Why are these patriarchal Roma leaders secretly envied by their Gadje partners when they occupy posts in unstable public bodies and face confused Roma and non-Roma voters – some of whom may even have been bribed?” (Gheorghe 2013: 159).

78

FROM “HUMAN DEVELOPMENT” TO “MULTIDIMENSIONAL POVERTY”

quential process. Agency is both an outcome and a determinant of poverty alleviation. Severe poverty lowers the level of aspiration, which makes even the limited opportunities an individual may have to a large extent hypothetical. In this way Roma are locked into a vicious cycle of exclusion that is reinforced by the very survival strategies that have evolved in response to severe poverty. These strategies are rooted in “low agency” social structures and accommodate to the existing status quo without challenging it. An obvious example is employment. Control over access to job creation programs and public works at the local level is a powerful source of leverage for local authorities and Roma community intermediaries, who often decide who is hired and who is not, thus reinforcing the patron-client relations that are already prevalent within the communities (Szalai 2013: 39-41). Developing an effective approach to agency would also require revisiting the conventional class-based analysis of Roma exclusion. Roma marginalization is not exclusively based on ethnicity, but shows clear characteristics of class division (UNDP 2002: 73). The disempowerment and exploitation of Roma communities, both from within and from outside cannot be addressed seriously, unless it is conceptualised as exploitation – regardless of how “outdated” this concept might sound. An additional difficulty in nurturing agency in Roma communities is the tendency of those few who manage to escape the yoke of marginalization to often prefer to distance themselves from the community, whose very existence is associated with deprivation and low social status. This “educated Roma flight” resembles the traditional “white flight,” and similarly contributes to those marginalized (those who remain in the slums) becoming even more deeply entrenched in marginalization (Szalai 2013: 27-28). The emancipation of Roma citizens would also meet the resistance from parts of the old Roma elites.43 Genuine support to agency in Roma communities would require breaking a number of taboos and stereotypes. It would entail much more than the “political participation” of Roma that too often consists of the incorporation of selected Roma activists into the Gadze power structures without giving them any real power and resources.44 It would also require redrawing traditional lines of loyalty and solidarity from the community and clan-based loyalty to broader civic approach. Keeping close, but distinct societal structures and not mingling with Gadzo has always been a constituent part of Roma 43/ Valeriu Nikolae is explicit in that regard: “The peer pressure at this moment in the Romani movement is against improvements and very much for keeping the status quo. Old Roma activists as well as the political elites do their utmost to bully into submission or to block any groups or initiatives that might undermine their positions or ideas. Criticism is confused with virulent personal attacks, labelling, smearing campaigns or threats” (Nikolae 2013). 44/ As the prominent Roma activist Željko Jovanović puts it, “Our elites, in their struggle for status, are controlled by far greater powers. Our communities, immersed in a daily struggle for survival, have been defined as target groups of beneficiaries and not recognised as a political constituency of rights-bearing citizens... the power we need is to enable us to assume responsibility for our own future. Those in authority have allowed us to participate but real power will never be given to us. Using every available democratic means we must take it ourselves” (Jovanović 2013: 369).

79

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

identity (Ivanov 2012: 80-85). Although this pattern may have been an effective strategy in the 18th and 19th century, it is no longer acceptable. Would a shift to an agency-focused paradigm entail redefinition of some fundamental elements of Roma identity? Definitely, however, it appears to be the only sustainable approach to real inclusion of Roma in their societies. Anything less would devolve into an endless process of inclusion that is never intended to be achieved. The real Roma communities would only continue to be involved as pawns, playing the role that they are given – of the marginalized and deprived, dependent on the tokens of support, but exploited from the outside and from within. Political participation is definitely one way to approach supporting agency. It would entail a lengthy process to overcome the separation of various intermediaries from the people in real communities and the resources devoted to improving their lives (economic and political). It would not be quick or easy, but unless it is done, the rhetoric of participation will remain a hollow electoral ritual (periodic casting of votes for parties that visit Roma communities only prior to elections). We are far removed from such an initiative – which is the major reason why the proposed multidimensional poverty measure is missing its core—agency—dimension. There is not much to measure yet, but hopefully there will be in the future. Otherwise, the most that the Roma inclusion process could achieve would be to meet some basic needs. Without agency the Roma would remain deprived of genuine human development.

80

6

COMPARING THE DIFFERENT APPROACHES

The evidence presented above highlights the differences between money- and people-centred measures of poverty. They should be borne in mind not just (and not primarily) for proper poverty monitoring, but more importantly for policy formulation. The conceptual framework on which a poverty analysis is based determines the responses.

Multidimensional versus money-metric measures The choice of metrics indeed yields different results. Figure 47 illustrates the different outcomes from the application of multidimensional versus monetary metrics as the differences in the trends of the two poverty estimates during 2004-2011. The most Figure 47: Monetary and multidimensional poverty rates of Roma, 2004-2011 (expenditure-based) (%)

Source: UNDP vulnerable groups survey 2004, UNDP/World Bank/EC regional Roma survey 2011.

81

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 48: Overlaps between absolute, relative and multidimensional poverty estimates (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

dramatic case is Albania, where the monetary poverty rate of Roma fell by more than half (from 78% to 36%), while multidimensional poverty rate increased from 49% to 66% over the same period. Similarly, in Serbia monetary poverty declined from 57% to 26%, while the multidimensional poverty rate increased from 51% to 55%. In the other countries the direction of change was the same (a decline in poverty measured by both metrics), but the magnitude of change differs in several countries. Multidimensional poverty rates among Roma declined more sharply than monetary poverty in Bosnia and Herzegovina, Montenegro, and FYR Macedonia; the decline was similar in both metrics in Croatia and Bulgaria; while the decline in monetary poverty in Romania was sharper than in multidimensional poverty. Figure 48, in contrast points to the weaknesses inherent in juxtaposing these two poverty measures. The same households and individuals fall into different poverty categories depending on the poverty concept and metric applied. As the left side of the Figure shows, 24% of Roma living at risk of marginalization are “relatively poor”, but not absolutely poor or multidimensional poor; while 6% are poor both relatively and absolutely. 19% are poor both relatively and multidimensionally. Only 17% are poor by all three definitions. The share of those who are poor only multidimensionally is 9%, while the share of those who are only absolutely poor is negligible. The share of people who are not poor by any of the criteria is 19%. If we consider as “poor” only those who are poor on all three counts, poverty rates “fall” to 17% for Roma and to 2% for their non-Roma neighbours—a large gap, but far from observable reality.

82

COMPARING THE DIFFERENT APPROACHES

The conceptual frameworks underlying different poverty metrics have critically important implications for poverty alleviation policies. The case of monetary poverty is indicative in that regard. Seeing poverty as a monetary phenomenon implicitly pushes policy responses toward “income maximization” approaches, leading to policies formulated primarily based on a “cost versus benefit” assessment. While costs and benefits undoubtedly must be integrated into any policy response, the exercise cannot be reduced to monetary costs and benefits only. A good case in point is the “costs of Roma exclusion” series of studies promoting the idea of Roma inclusion as “smart economics” (De Laat and Bodewig, 2011). Their goal—providing policy makers with arguments (ideally, numeric) in favour of inclusion efforts—is honourable and a debate on the benefits of inclusion is arguably needed. However, many of the costs and benefits of exclusion are not in the sphere of economics. The “Roma inclusion as smart economics” approach estimates the direct and indirect economic losses to society resulting from the exclusion of a large share of the population from employment.45 These losses include lost tax revenue and higher social security payments (e.g., unemployment benefits, social assistance), as well as the higher costs of addressing the asocial behaviour associated with long-term unemployment (e.g., petty crime). The analysis uses a partial equilibrium model to estimate desirable changes in tax revenues, social assistance payments, etc., that would result from increases in Roma employment. On that basis the cost of Roma exclusion is estimated on average at more than 3% of GDP lost per year in Slovakia, Serbia and the Czech Republic. The problem with such models (as with any model) lies in their assumptions. The study implicitly assumes infinite employment opportunities for Roma, thereby reducing the complexity of Roma labour market exclusion by attributing it solely to discriminatory factors, prejudice and insufficient qualification. While these factors are clearly present, they are not the sole (and perhaps not even the major) drivers of Roma labour market exclusion. These models implicitly assume either the existence of unoccupied jobs in the localities where Roma live (who “just” need to be employed), or of sufficient mobility of Roma workers, either up the professional qualification ladder (from less to more skilled occupations and services) or territorially (i.e., moving from areas of high to low unemployment rates). In reality both assumptions are often wrong. Local labour markets in Central and Southeast Europe are highly fragmented and stratified. “Lump of labour” theories are more appropriate for describing their dynamics. The vertical mobility of Roma is often marginal, and their territorial mobility is taking place primarily across countries—with Roma emigration producing backlash in the receiving societies (Cherkezova and Tomova 2013). Roma inclusion strategies based on “smart economics” may therefore produce more questions than answers. If inclusion of Roma is economically smart, why have businesses not taken advantage of these opportunities? This again reminds us that the concepts on which policies and metrics are based really matter and that the benefits 45/ The logic of “Roma inclusion as smart economics” is much more complicated and deserves a separate study going beyond the scope of the current research.

83

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure 49: Overlaps between material deprivation and multidimensional poverty estimates, 2011 (%)

Source: UNDP/World Bank/EC regional Roma survey 2011.

of Roma inclusion are less likely to be found in the monetary than in the social, human development domain. Roma inclusion is a matter of citizenship, equal opportunity, individual freedom, and fundamental rights. Benefit to the society stems from the realization of the guarantee that the rights of every single citizen—Roma and non-Roma alike—will be protected and respected. The Roma are an extreme case in this regard, which simply makes the argument more explicit. Roma inclusion has other (social, noneconomic, intangible, non-monetary) benefits that cannot be easily quantified in economic or econometric models. Accordingly, Roma poverty definition and monitoring cannot, and should not, be reduced to monetary poverty.

Other estimates of well-being While the other metrics of well-being discussed above are a step toward grasping the complexity of Roma poverty, they still fall short of completing the picture. Figure 49 illustrates the results of the two non-monetary approaches to identifying and quantifying poverty (MPI and deprivation). The first shows the share of individuals who are considered “poor” using all three approaches (absolute, relative, and multidimensional poverty). The second shows the same in terms of material deprivation and multidimensional poverty. The degree of overlap is greatest between multidimensional poverty

84

COMPARING THE DIFFERENT APPROACHES

and material deprivation. One might say “this is because material deprivation is included in the MPI”. But the former constitutes only 1/16th of the latter. These comparisons raise a difficult question: which of all these approaches is the “right” one? The answer is: “it depends on the purpose of the exercise”. If the intention is to highlight the differences in poverty levels, the simplest approach (monetary poverty estimates) is sufficient. However, it is only possible to understand and address the roots of Roma poverty when they are conceptualized from a human development perspective. This requires an analysis that goes beyond monetary estimates and the deconstruction of multidimensional poverty to identifying individual problem zones that require targeted, sector-specific interventions.

85

7

CONCLUSIONS The analysis presented here leads to several important conclusions regarding the specifics of Roma poverty and the relevance of different approaches to its measurement and monitoring.

A wide range of methodologies for measuring poverty exists. The analysis presented above highlights the differences between money-metric and people-centred measures of poverty. They should be borne in mind not just (and not primarily) for proper poverty monitoring, but more importantly for policy formulation. The conceptual basis of a poverty analysis determines the policy responses. If the ultimate purpose is to better understand the underlying mechanisms and the factors determining poverty—and to help overcome them—the choice of approach should be guided by policy priorities. The complex specifics of Roma poverty require multidimensional approaches. The proposed multi-dimensional poverty index is one example of such an approach. It is better suited for tracking the poverty of marginalized and excluded groups such as the Roma and can be applied at negligible additional cost. While there are areas of overlap, the multidimensional poverty index shows some differences in trend compared to purely monetary measures. As noted above, different outcomes were estimated for 2004-2011. The most dramatic is the case of Albania where the monetary poverty rate of Roma was more than cut in half (from 78% to 36%) while multidimensional poverty rate increased from 49% to 66%. Likewise, in Serbia monetary poverty declined from 57% to 26%, while the multidimensional poverty rate increased from 51% to 55%. They also result in different policy recommendations. For example, monetary poverty only allows one to target income or expenditures as a way to increase this measure, whereas, the “policy graph” (Figure 35) and the probit results (Table 4), which are based on the multidimensional poverty index, show the wide array of targets for potential improvements to Roma exclusion. Although the proposed multidimensional poverty index is better suited for tracking the poverty of Roma, it is still not perfect. Similarly to all other poverty measures, it does not cover the entire spectrum of human-centred aspects of poverty. A major element—agency—is missing. The individual capabilities and the freedom to take matters into one’s own hands are a crucial dimension and driver of Roma exclusion. Apart from the ethnic dimensions, exploitation and cross-generational replication of poverty also connotate elements of class conflict. All these must be recognized for the picture (and the resulting policies) to be accurate. Thus expanding the multidimensional poverty measures to include indicators for agency will be necessary to make the index even more relevant and effective policy tool.

86

The better reflection of local specifics and intra-group dynamics is a second lacuna. Aggregate figures are important—but to be properly understood, they need to be examined in a concrete local context. Qualitative data is a key element in developing such a deeper micro-level understanding. Socio-economic statistics matter (and there are a lot of such data). However, to effectively support social and economic change, a more comprehensive picture of the values and behavioural patterns of individuals within the context in which they take their daily decisions is required. Thus far, a great deal of knowledge has been accumulated, but there are not sufficient data that can be correlated with socio-economic indicators. All of these areas require further hard work if the Roma are to enjoy genuine human development.

87

7

ANNEXES Bibliography

Alkire, Sabina (2008). Concepts and Measures of Agency. Queen Elizabeth House, University of Oxford, OPHI Working Paper 010. Alkire, Sabina (2008), “Choosing Dimensions: The Capability Approach and Multidimensional Poverty”, in: Kakwani, N. and Silber, J. (2008). The Many Dimensions of Poverty, New York: Palgrave Macmillan, pp. 89-119. Alkire, Sabina (2010). Human Development: Definitions, Critiques, and Related Concepts. Background paper for the 2010 Human Development Report. Alkire, Sabina; Foster, James (2011). Understandings and Misunderstandings of Multidimensional Poverty MeasurementOPHI Working Paper No. 43. Oxford: University of Oxford. Atkinson, Anthony; Bourguignon, Francois (1999). Poverty and Inclusion from a World Perspective. Paper prepared for Annual Bank Conference on Development Economics in Europe Proceedings: 1999-2000. Paris: Documentation Française. Bíró, András (2013). The Price of Roma Integration. In: Bíró, András; Gheorghe, Nicolae; Kovats, Martin et al. (2013). From victimhood to citizenship. Kossuth Kiadó. Brüggemann, Christian (2012). Roma Education in Comparative Perspective. Analysis of the UNDP/World Bank/EC Regional Roma Survey 2011. Roma Inclusion Working Papers. UNDP: Bratislava. Cauce, Ana Mari; Gordon, Edmund W. (2013). Toward the Measurement of Human Agency and the Disposition to Express It. Working paper, The Gordon Commission on the Future of Assessment in Education. Chen, Shaohua; Ravallion, Martin (2007). Absolute Poverty Measures for the Developing World, 1981-2004. Policy Research Working Paper 4211. World Bank: Washington, D.C. Coudouel, Aline, Jesko Hentschel and Quentin Wodon (2002), Poverty Measurement and Analysis. World Bank: Washington D.C. Cukrowska, Ewa; Kóczé, Angela (2013). Interplay between gender and ethnicity: exposing structural disparities of Romani women. Analysis of the UNDP/World Bank/EC regional Roma survey data. Roma Inclusion Working Papers. UNDP: Bratislava. De Laat, Joost; Bodewig, Christian (2011). Roma Inclusion is Smart Economics - Illustrations from Bulgaria, Czech Republic, Romania and Serbia. World Bank, ECA Knowledge Brief 39, April 2011.

88

ERRC (2012). Fact Sheet: Overrepresentation of Romani Children in Special Education in Macedonia. ERRC: Budapest. http://www.errc.org/cms/upload/file/macedonia-factsheet-education-en-30-august-2012.pdf ERRC (2014). A Long Way to Go. Overrepresentation of Romani Children in “Special Schools” in Serbia. A Report by the European Roma Rights Centre. ERRC: Budapest. http://www.errc.org/cms/upload/file/serbia-education-report-a-long-way-to-go-serbian-13-march-2014.pdf FRA (2014, forthcoming). Roma Pilot Survey – Technical report: methodology, sampling and fieldwork. Publications office of the European Union: Luxemburg. FRA (2012). The situation of Roma in 11 EU Member States - Survey results at a glance. Publications office of the European Union: Luxemburg. FRA (2014, forthcoming). The situation of Roma in 11 EU Member States – Employment. Data in focus. Publications office of the European Union: Luxemburg. FRA (2014, forthcoming). The situation of Roma in 11 EU Member States – Education. Data in focus. Publications office of the European Union: Luxemburg. Garroway, Christopher; de Laiglesia, Juan R. (2012). On the relevance of relative poverty for developing countries. OECD Development Centre Working Paper 314. OECD: Paris. Gheorghe, Nicolae; Pulay, Gergő (2013). Choices to Be Made And Prices To Be Paid: Potential Roles And Consequences In Roma Activism And Policy-Making. In: Bíró, András; Gheorghe, Nicolae; Kovats, Martin et al. (2013). From victimhood to citizenship. Kossuth Kiadó. Gordon, David (2002). International measurements of poverty and anti-poverty policies. In: Townsend, Peter and Gordon, David (eds.). World Poverty: New Policies to Defeat an Old Enemy. Bristol: Polity Press. Ibrahim, Solava; Alkire, Sabina (2007). Agency & Empowerment: A proposal for internationally comparable indicators. Queen Elizabeth House, University of Oxford, OPHI Working Paper 05. ILO (1976). Employment, Growth and Basic Needs: A One-World Problem. ILO: Geneva. Ivanov Andrey; Kling, Jaroslav; Kagin, Justin (2012). Integrated household surveys among Roma populations: one possible approach to sampling used in the UNDP-World Bank-EC Regional Roma Survey 2011. Roma Inclusion Working Papers. UNDP: Bratislava. Ivanov, Andrey (2012). Quantifying the Unquantifiable: Defining Roma Populations in Quantitative Surveys. Население Issue no.3-4 /2012, pp. 79-95. Jovanović, Željko (2013). Values, leadership, power. In: Bíró, András; Gheorghe, Nicolae; Kovats, Martin et al. (2013). From victimhood to citizenship. Kossuth Kiadó. Kagin, Justin (2013). Norms, values and the effect of churches on an excluded community:

89

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

The case of the Roma in Eastern Europe. Dissertation submitted in partial satisfaction of the requirements for the degree of Doctor of Philosophy, University of California, Davis. Kakwani, Nanak; Son, Hyun H. (2006). New Global Poverty Counts, Working paper no. 29. UNDP International Poverty Centre: Brasil. Kóczé, Angela (2012). Civil society, civil involvement and social inclusion of the Roma. Roma Inclusion Working Papers. Bratislava: UNDP. Kóczé, Angela; Rövid, Marton (2012) “Pro-Roma global civil society: acting for, with or instead of Roma?” in Mary Kaldor, Henrietta L. Moore and Sabine Selchow, Global Civil Society 2012: Ten Years of Critical Reflection, Basingstoke: Palgrave Macmillan. 2012. pp. 110-122. Kovats, Martin (2013). Integration and The Politicisation Of Roma Identity. In: Bíró, András; Gheorghe, Nicolae; Kovats, Martin et al. (2013). From victimhood to citizenship. Kossuth Kiadó. Lister, Ruth (2004). Poverty. Polity Press: Cambridge. Metodieva Maria (coordinator), Abramova, Svetlana Belcheva, Dragomir (2012). Beyond the myths and prejudices: Roma in Bulgaria. EU inclusive - data transfer and exchange of good practices between Romania, Bulgaria, Spain and Italy concerning inclusive of the Roma population.. Editura Dobrogea: Open Society. Mihailov, Dotcho (2012). The health situation of Roma communities: Analysis of the data from the UNDP/World Bank/EC Regional Roma Survey 2011. Roma Inclusion Working Papers. UNDP: Bratislava. Nikolae, Valeriu (2013). On Roma civil society and donors. Blog on “The WorldPress. com,” http://valeriucnicolae.wordpress.com/2013/05/ O’Higgins, Niall (2012). Roma and non-Roma in the Labour Market in Central and South Eastern Europe. Roma Inclusion Working Papers. UNDP: Bratislava. Perić, Tatjana (2012). The Housing Situation of Roma Communities: Regional Roma Survey 2011. Roma Inclusion Working Papers. UNDP: Bratislava. Ravallion, Martin (1992). Poverty Comparisons. A guide to concepts and methods. LSMS Working Paper 88. The World Bank: Washington D.C. Ravallion, Martin (2003) The Debate on Globalization, Poverty and Inequality: Why Measurement Matter?, Policy Research Working Paper 3038, World Bank: Washington D.C. Ravallion, Martin (2010). Poverty lines across the world. Policy research Working Paper 5284. The World Bank: Washington D.C. Ravallion, Martin; Bidani, Benu (1994). How robust is a poverty profile? The World bank Economic Review, Vol. 8, No. 1. Ravallion, Martin; Chen, Shaohua; Sangraula, Prem (2008). Dollar a Day Revisited. The World Bank Development Research Group, Policy Research Working Paper, 4620. World Bank: Washington D.C.

90

ANNEXES

Ruggeri Laderchi, Caterina; Saith, Ruhi; Stewart; Frances (2003). Does it Matter that we do not agree on the Definition of Poverty? A Comparison of Four Approaches. QEH Working Paper Series, Working Papers 107. Oxford Development Studies. Samman, Emma; Santos, Maria Emma (2009). Agency and Empowerment: A review of concepts, indicators and empirical evidence. Queen Elizabeth House, University of Oxford, OPHI Working Paper Sen, Amartya (1985) Well-being, agency, and freedom. The Dewey Lectures 1984. The Journal of Philosophy, 82, 169-221. Sen, Amartya (1987). Commodities and Capabilities. Amsterdam: North-Holland. Sen, Amartya (2000a). Development as freedom. Anchor books: New York. Sen, Amartya (2000b). Social exclusion: concept, application, and scrutiny. Office of Environment and Social Development Asian Development Bank: Manila. Szalai, Julia (2013). Faces and causes of Roma marginalization in local communities. A comparative report of the research in Phase 2 of the OSI-funded research on local level contextualization of UNDP 2011 survey data. Unpublished manuscript available from www.europeandcis.undp.org. Tarnovschi, Daniela et al. (2012). Roma from Romania, Bulgaria, Italy and Spain between Social Inclusion and Migration: comparative study. Constanta: Editura Dobrogea. The World Bank (1990). Poverty. World Development Report . Oxford: Oxford University Press The World Bank (2000). Attacking Poverty. World Development Report 2000/2001. World Bank and Oxford University Press. Washington, DC: World Bank. Thorbecke, Erik (2005). Multi-dimensional poverty: Conceptual and measurement issues. International conference “The many dimensions of poverty.” UNDP IPC: Brasilia. Townsend, Peter (1962). The Meaning of Poverty. The British Journal of Sociology, Vol. 13, No. 3 (Sep., 1962), pp. 210-227. UNDP (2002). The Roma in Cenytral and Eastern Europe: Avoiding the Dependency Trap. A regional human development report. UNDP: Bratislava. UNDP (2010). Human Development Report 2011. The Real Wealth of Nations: Pathways to Human Development. UNDP: New York. UNDP (1997). Human Development to Eradicate Poverty: Human Development Report. UNDP: New York. UNDP (2006). At Risk: Roma and Displaced in South-East Europe. Regional human development report. UNDP: Bratislava. UNDP (2011) Beyond Transition – Towards Inclusive Societies. Regional human development report. UNDP: Bratislava.

91

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Contry profiles Figure A: Average years spent in education by poverty status for the age 25-64

92

ANNEXES

Figure B: Share of the unemployed as a percentage of those in the labour force (15-64)

93

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Figure C: Share of employed people who do not have a written contract (ages 15-64)

94

ANNEXES

Figure D: Share of respondents who said that it is fully acceptable stealing food if a family goes hungry

95

ROMA POVERTY FROM A HUMAN DEVELOPMENT PERSPECTIVE

Other publications in the “Roma Inclusion Working Papers” series:

Brüggemann, Christian (2012). Roma Education in Comparative Perspective. Analysis of the UNDP/World Bank/EC Regional Roma Survey 2011. Roma Inclusion Working Papers. Bratislava: UNDP. Cherkezova, Stoyanka; Tomova, Ilona (2013). An option of last resort? Migration of Roma and Non-Roma from CEE Countries. Roma Inclusion Working Papers. Bratislava: UNDP. Cukrowska, Ewa; Kóczé, Angela (2013). Interplay between gender and ethnicity: exposing structural disparities of Romani women. Analysis of the UNDP/World Bank/EC regional Roma survey data. Roma Inclusion Working Papers. Bratislava: UNDP. Hurrle, Jakob; Ivanov, Andrey; Grill, Jan; Kling, Jaroslav; Škobla, Daniel (2012). Uncertain impact: Have the Roma in Slovakia benefitted from the European Social Fund? Findings from an Analysis of ESF Employment and Social Inclusion Projects in the 2007 – 2013 Programming Period. Roma Inclusion Working Papers. Bratislava: UNDP. Ivanov, Andrey; Kling, Jroslav; Kagin, Justin (2012). Integrated household surveys among Roma populations: one possible approach to sampling used in the UNDPWorld Bank-EC Regional Roma Survey 2011. Roma Inclusion Working Papers. Bratislava: UNDP Kóczé, Angela (2012). Civil society, civil involvement and social inclusion of the Roma. Roma Inclusion Working Papers. Bratislava: UNDP Mihailov, Dotcho (2012). The health situation of Roma communities: Analysis of the data from the UNDP/World Bank/EC Regional Roma Survey 2011. Roma Inclusion Working Papers. Bratislava: UNDP. O’Higgins, Niall (2012). Roma and non-Roma in the Labour Market in Central and South Eastern Europe. Roma Inclusion Working Papers. Bratislava: UNDP. Perić, Tatjana (2012). The Housing Situation of Roma Communities: Regional Roma Survey 2011. Roma Inclusion Working Papers. Bratislava: UNDP.

96

UNDP Regional Support Centre for Europe and CIS, Istanbul http://europeandcis.undp.org